chore: import upstream snapshot with attribution
This commit is contained in:
wehub-resource-sync
@@ -0,0 +1,5 @@
|
||||
{
|
||||
"<h1>GPT-NeoX</h1>\n<p>This is a simple implementation of <a href=\"https://arxiv.org/abs/2204.06745\">Eleuther GPT-NeoX</a> for inference and fine-tuning.</p>\n<ul><li><a href=\"model.html\">Model definition</a> </li>\n<li><a href=\"tokenizer.html\">Tokenizer</a> </li>\n<li><a href=\"checkpoint.html\">Checkpoint downloading and loading helpers</a> </li>\n<li><a href=\"utils/index.html\">Utilities</a> </li>\n<li><a href=\"utils/llm_int8.html\">LLM.int8() quantization</a></li></ul>\n<h3><a href=\"samples/__init__.py\">Samples</a></h3>\n<ul><li><a href=\"samples/generate.html\">Generating text</a> </li>\n<li><a href=\"samples/finetune.html\">Fine-tuning the biases with pipeline-parallel</a> </li>\n<li><a href=\"samples/llm_int8.html\">Generating text with LLM.int8()</a></li></ul>\n<h3><a href=\"evaluation/__init__.py\">Evaluation</a></h3>\n<ul><li><a href=\"evaluation/half_precision.html\">Evaluating half precision model on a single GPU</a> </li>\n<li><a href=\"evaluation/llm_int8.html\">Evaluating LLM.int8() model</a></li></ul>\n<p><strong>Official <a href=\"https://www.eleuther.ai\">Eleuther</a> GPT-NoeX is source code is available at <a href=\"https://github.com/eleutherai/gpt-neox\">eleutherai/gpt-neox</a>.</strong></p>\n": "<h1>GPT \u30cd\u30aa\u30c3\u30af\u30b9</h1>\n<p>\u3053\u308c\u306f\u3001<a href=\"https://arxiv.org/abs/2204.06745\">\u63a8\u8ad6\u3068\u5fae\u8abf\u6574\u306e\u305f\u3081\u306eEleuther GPT-Neox\u306e\u7c21\u5358\u306a\u5b9f\u88c5\u3067\u3059</a>\u3002</p>\n<ul><li><a href=\"model.html\">\u30e2\u30c7\u30eb\u5b9a\u7fa9</a></li>\n<li><a href=\"tokenizer.html\">\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc</a></li>\n<li><a href=\"checkpoint.html\">\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u306e\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3068\u8aad\u307f\u8fbc\u307f\u30d8\u30eb\u30d1\u30fc</a></li>\n<li><a href=\"utils/index.html\">\u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3</a></li>\n<li><a href=\"utils/llm_int8.html\">llm.int8 () \u91cf\u5b50\u5316</a></li></ul>\n<h3><a href=\"samples/__init__.py\">[\u30b5\u30f3\u30d7\u30eb]</a></h3>\n<ul><li><a href=\"samples/generate.html\">\u30c6\u30ad\u30b9\u30c8\u306e\u751f\u6210</a></li>\n<li><a href=\"samples/finetune.html\">\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u30d1\u30e9\u30ec\u30eb\u306b\u3088\u308b\u30d0\u30a4\u30a2\u30b9\u306e\u5fae\u8abf\u6574</a></li>\n<li><a href=\"samples/llm_int8.html\">LLM.int8 () \u3092\u4f7f\u7528\u3057\u3066\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210\u3059\u308b</a></li></ul>\n<h3><a href=\"evaluation/__init__.py\">\u8a55\u4fa1</a></h3>\n<ul><li><a href=\"evaluation/half_precision.html\">\u5358\u4e00\u306e GPU \u3067\u306e\u534a\u7cbe\u5ea6\u30e2\u30c7\u30eb\u306e\u8a55\u4fa1</a></li>\n<li><a href=\"evaluation/llm_int8.html\">LLM.int8 () \u30e2\u30c7\u30eb\u306e\u8a55\u4fa1\u4e2d</a></li></ul>\n<p><strong><a href=\"https://www.eleuther.ai\">Eleuther GPT-noex\u306e\u516c\u5f0f\u30bd\u30fc\u30b9\u30b3\u30fc\u30c9\u306feleutherai/gpt-neox\u3067\u5165\u624b\u3067\u304d\u307e\u3059</a><a href=\"https://github.com/eleutherai/gpt-neox\">\u3002</a></strong></p>\n",
|
||||
"GPT-NeoX": "GPT \u30cd\u30aa\u30c3\u30af\u30b9",
|
||||
"Simple GPT-NeoX implementation": "\u30b7\u30f3\u30d7\u30eb\u306a GPT \u30cd\u30aa\u30c3\u30af\u30b9\u306e\u5b9f\u88c5"
|
||||
}
|
||||
@@ -0,0 +1,5 @@
|
||||
{
|
||||
"<h1>GPT-NeoX</h1>\n<p>This is a simple implementation of <a href=\"https://arxiv.org/abs/2204.06745\">Eleuther GPT-NeoX</a> for inference and fine-tuning.</p>\n<ul><li><a href=\"model.html\">Model definition</a> </li>\n<li><a href=\"tokenizer.html\">Tokenizer</a> </li>\n<li><a href=\"checkpoint.html\">Checkpoint downloading and loading helpers</a> </li>\n<li><a href=\"utils/index.html\">Utilities</a> </li>\n<li><a href=\"utils/llm_int8.html\">LLM.int8() quantization</a></li></ul>\n<h3><a href=\"samples/__init__.py\">Samples</a></h3>\n<ul><li><a href=\"samples/generate.html\">Generating text</a> </li>\n<li><a href=\"samples/finetune.html\">Fine-tuning the biases with pipeline-parallel</a> </li>\n<li><a href=\"samples/llm_int8.html\">Generating text with LLM.int8()</a></li></ul>\n<h3><a href=\"evaluation/__init__.py\">Evaluation</a></h3>\n<ul><li><a href=\"evaluation/half_precision.html\">Evaluating half precision model on a single GPU</a> </li>\n<li><a href=\"evaluation/llm_int8.html\">Evaluating LLM.int8() model</a></li></ul>\n<p><strong>Official <a href=\"https://www.eleuther.ai\">Eleuther</a> GPT-NoeX is source code is available at <a href=\"https://github.com/eleutherai/gpt-neox\">eleutherai/gpt-neox</a>.</strong></p>\n": "<h1>\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca</h1>\n<p>\u0db8\u0dd9\u0dba\u0d85\u0db1\u0dd4\u0db8\u0dcf\u0db1\u0dba \u0dc3\u0dc4 \u0db8\u0db1\u0dcf\u0dc0 \u0dc3\u0dd4\u0dc3\u0dbb \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf <a href=\"https://arxiv.org/abs/2204.06745\">\u0d91\u0dbd\u0dd2\u0dba\u0dd4\u0dad\u0dbb\u0dca \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca</a> \u0dc3\u0dbb\u0dbd \u0dbd\u0dd9\u0dc3 \u0d9a\u0dca\u0dbb\u0dd2\u0dba\u0dcf\u0dad\u0dca\u0db8\u0d9a \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0d9a\u0dd2. </p>\n<ul><li><a href=\"model.html\">\u0d86\u0daf\u0dbb\u0dca\u0dc1 \u0d85\u0dbb\u0dca\u0dae \u0daf\u0dd0\u0d9a\u0dca\u0dc0\u0dd3\u0db8</a> </li>\n<li><a href=\"tokenizer.html\">\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca</a> </li>\n<li><a href=\"checkpoint.html\">\u0db4\u0dd2\u0dbb\u0dd2\u0d9a\u0dca\u0dc3\u0dd4\u0db8\u0dca \u0dbd\u0d9a\u0dca\u0dc2\u0dca\u0dba\u0dba \u0db6\u0dcf\u0d9c\u0dad \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0dc4 \u0db4\u0dd0\u0da7\u0dc0\u0dd3\u0db8 \u0dc3\u0dc4\u0dcf\u0dba\u0d9a\u0dba\u0dd2\u0db1\u0dca</a> </li>\n<li><a href=\"utils/index.html\">\u0d8b\u0db4\u0dba\u0ddd\u0d9c\u0dd2\u0dad\u0dcf</a> </li>\n<li><a href=\"utils/llm_int8.html\">LLM.INT8 () \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba</a></li></ul>\n<h3><a href=\"samples/__init__.py\">\u0dc3\u0dcf\u0db8\u0dca\u0db4\u0dbd</a></h3>\n<ul><li><a href=\"samples/generate.html\">\u0db4\u0dd9\u0dc5 \u0d8b\u0dad\u0dca\u0db4\u0dcf\u0daf\u0db1\u0dba</a> </li>\n<li><a href=\"samples/finetune.html\">\u0db1\u0dbd \u0db8\u0dcf\u0dbb\u0dca\u0d9c\u0dba\u0dda \u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb\u0dc0 \u0d87\u0dad\u0dd2 \u0d85\u0d9c\u0dad\u0dd3\u0db1\u0dca \u0db8\u0db1\u0dcf\u0dc0 \u0dc3\u0d9a\u0dc3\u0dca \u0d9a\u0dd2\u0dbb\u0dd3\u0db8</a> </li>\n<li><a href=\"samples/llm_int8.html\">LLM.INT8 () \u0dc3\u0db8\u0d9f \u0db4\u0dd9\u0dc5 \u0da2\u0db1\u0db1\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8</a></li></ul>\n<h3><a href=\"evaluation/__init__.py\">\u0d87\u0d9c\u0dba\u0dd3\u0db8</a></h3>\n<ul><li><a href=\"evaluation/half_precision.html\">\u0dad\u0db1\u0dd2 GPU \u0db8\u0dad \u0d85\u0dbb\u0dca\u0db0 \u0db1\u0dd2\u0dbb\u0dc0\u0daf\u0dca\u0dba\u0dad\u0dcf \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0d9a\u0dca \u0d87\u0d9c\u0dba\u0dd3\u0db8</a> </li>\n<li><a href=\"evaluation/llm_int8.html\">LLM.INT8 () \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0d87\u0d9c\u0dba\u0dd3\u0db8</a></li></ul>\n<p><strong>\u0db1\u0dd2\u0dbd <a href=\"https://www.eleuther.ai\">\u0d91\u0dbd\u0dd2\u0dad\u0dbb\u0dca</a> \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0ddc\u0d9a\u0dca\u0dc3\u0dca \u0dba\u0db1\u0dd4 \u0db4\u0dca\u0dbb\u0db7\u0dc0 \u0d9a\u0dda\u0dad\u0dba <a href=\"https://github.com/eleutherai/gpt-neox\">eleutherai/gpt-neox</a> \u0dc0\u0dbd\u0dd2\u0db1\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0dad \u0dc4\u0dd0\u0d9a\u0dd2\u0dba. </strong></p>\n",
|
||||
"GPT-NeoX": "\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca",
|
||||
"Simple GPT-NeoX implementation": "\u0dc3\u0dbb\u0dbd \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0d9a\u0dca\u0dbb\u0dd2\u0dba\u0dcf\u0dad\u0dca\u0db8\u0d9a \u0d9a\u0dd2\u0dbb\u0dd3\u0db8"
|
||||
}
|
||||
@@ -0,0 +1,5 @@
|
||||
{
|
||||
"<h1>GPT-NeoX</h1>\n<p>This is a simple implementation of <a href=\"https://arxiv.org/abs/2204.06745\">Eleuther GPT-NeoX</a> for inference and fine-tuning.</p>\n<ul><li><a href=\"model.html\">Model definition</a> </li>\n<li><a href=\"tokenizer.html\">Tokenizer</a> </li>\n<li><a href=\"checkpoint.html\">Checkpoint downloading and loading helpers</a> </li>\n<li><a href=\"utils/index.html\">Utilities</a> </li>\n<li><a href=\"utils/llm_int8.html\">LLM.int8() quantization</a></li></ul>\n<h3><a href=\"samples/__init__.py\">Samples</a></h3>\n<ul><li><a href=\"samples/generate.html\">Generating text</a> </li>\n<li><a href=\"samples/finetune.html\">Fine-tuning the biases with pipeline-parallel</a> </li>\n<li><a href=\"samples/llm_int8.html\">Generating text with LLM.int8()</a></li></ul>\n<h3><a href=\"evaluation/__init__.py\">Evaluation</a></h3>\n<ul><li><a href=\"evaluation/half_precision.html\">Evaluating half precision model on a single GPU</a> </li>\n<li><a href=\"evaluation/llm_int8.html\">Evaluating LLM.int8() model</a></li></ul>\n<p><strong>Official <a href=\"https://www.eleuther.ai\">Eleuther</a> GPT-NoeX is source code is available at <a href=\"https://github.com/eleutherai/gpt-neox\">eleutherai/gpt-neox</a>.</strong></p>\n": "<h1>GPT-neox</h1>\n<p>\u8fd9\u662f Ele <a href=\"https://arxiv.org/abs/2204.06745\">uther GPT-NEOX</a> \u7684\u7b80\u5355\u5b9e\u73b0\uff0c\u7528\u4e8e\u63a8\u7406\u548c\u5fae\u8c03\u3002</p>\n<ul><li><a href=\"model.html\">\u578b\u53f7\u5b9a\u4e49</a></li>\n<li><a href=\"tokenizer.html\">\u5206\u8bcd\u5668</a></li>\n<li><a href=\"checkpoint.html\">\u68c0\u67e5\u70b9\u4e0b\u8f7d\u548c\u52a0\u8f7d\u52a9\u624b</a></li>\n<li><a href=\"utils/index.html\">\u516c\u5171\u4e8b\u4e1a</a></li>\n<li><a href=\"utils/llm_int8.html\">llm.int8 () \u91cf\u5316</a></li></ul>\n<h3><a href=\"samples/__init__.py\">\u6837\u54c1</a></h3>\n<ul><li><a href=\"samples/generate.html\">\u751f\u6210\u6587\u672c</a></li>\n<li><a href=\"samples/finetune.html\">\u4f7f\u7528\u7ba1\u9053\u5e73\u884c\u5fae\u8c03\u504f\u5dee</a></li>\n<li><a href=\"samples/llm_int8.html\">\u4f7f\u7528 llm.int8 () \u751f\u6210\u6587\u672c</a></li></ul>\n<h3><a href=\"evaluation/__init__.py\">\u8bc4\u4f30</a></h3>\n<li><a href=\"evaluation/half_precision.html\">\u5728\u5355\u4e2a GPU \u4e0a\u8bc4\u4f30\u534a\u7cbe\u5ea6\u6a21\u578b</a></li> <ul>\n<li><a href=\"evaluation/llm_int8.html\">\u6b63\u5728\u8bc4\u4f30 llm.int8 () \u6a21\u578b</a></li></ul>\n<p><strong>\u5b98\u65b9\u7684 <a href=\"https://www.eleuther.ai\">Eleuther</a> GPT-NOEX \u662f\u6e90\u4ee3\u7801\u53ef\u5728 <a href=\"https://github.com/eleutherai/gpt-neox\">eleutherai/gpt-neox</a> \u83b7\u5f97\u3002</strong></p>\n",
|
||||
"GPT-NeoX": "GPT-neox",
|
||||
"Simple GPT-NeoX implementation": "\u7b80\u5355\u7684 GPT-NEOX \u5b9e\u73b0"
|
||||
}
|
||||
@@ -0,0 +1,25 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Checkpoints</h1>\n": "<h1>GPT-\u30cd\u30aa\u30c3\u30af\u30b9\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8</h1>\n",
|
||||
"<h2>Download all checkpoint files</h2>\n": "<h2>\u3059\u3079\u3066\u306e\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u30d5\u30a1\u30a4\u30eb\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9</h2>\n",
|
||||
"<h3>Get files to download</h3>\n<ul><p><em>Returns</em> a list of files to be downloaded</p></ul>\n": "<h3>\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3059\u308b\u30d5\u30a1\u30a4\u30eb\u3092\u53d6\u5f97</h3>\n<ul><p><em>\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3059\u308b\u30d5\u30a1\u30a4\u30eb\u306e\u30ea\u30b9\u30c8\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h3>Load a pair of checkpoint files</h3>\n<ul><li><span translate=no>_^_0_^_</span> pair of files to load </li>\n<p><em>Returns</em> the loaded parameter tensors</p></ul>\n": "<h3>1 \u7d44\u306e\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u30d5\u30a1\u30a4\u30eb\u3092\u30ed\u30fc\u30c9\u3059\u308b</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30ed\u30fc\u30c9\u3059\u308b\u30d5\u30a1\u30a4\u30eb\u306e\u30da\u30a2</li>\n</ul><p><em>\u30ed\u30fc\u30c9\u3055\u308c\u305f\u30d1\u30e9\u30e1\u30fc\u30bf\u30c6\u30f3\u30bd\u30eb\u3092\u8fd4\u3057\u307e\u3059</em>\u3002</p>\n",
|
||||
"<h3>Load a parameter by merging the partitions along first dimension</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>1 \u756a\u76ee\u306e\u6b21\u5143\u306b\u6cbf\u3063\u3066\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u3092\u30de\u30fc\u30b8\u3057\u3066\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u30ed\u30fc\u30c9\u3057\u307e\u3059</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u540d\u524d\u3067\u3059</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c 1 \u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li>\n<li><span translate=no>_^_3_^_</span>2 \u756a\u76ee\u306e\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li></ul>\n",
|
||||
"<h3>Load a parameter by merging the partitions along second dimension</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>2 \u756a\u76ee\u306e\u6b21\u5143\u306b\u6cbf\u3063\u3066\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u3092\u30de\u30fc\u30b8\u3057\u3066\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u30ed\u30fc\u30c9\u3057\u307e\u3059</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u540d\u524d\u3067\u3059</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c 1 \u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li>\n<li><span translate=no>_^_3_^_</span>2 \u756a\u76ee\u306e\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li></ul>\n",
|
||||
"<h3>Load an un-partitioned parameter</h3>\n<p>This does a sanity check to make use both partitions are the same</p>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u5316\u3055\u308c\u3066\u3044\u306a\u3044\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u30ed\u30fc\u30c9\u3059\u308b</h3>\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u4e21\u65b9\u306e\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u304c\u540c\u3058\u3067\u3042\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3059\u308b\u305f\u3081\u306e\u30b5\u30cb\u30c6\u30a3\u30c1\u30a7\u30c3\u30af\u304c\u884c\u308f\u308c\u307e\u3059\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u540d\u524d\u3067\u3059</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c 1 \u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li>\n<li><span translate=no>_^_3_^_</span>2 \u756a\u76ee\u306e\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li></ul>\n",
|
||||
"<h3>Load biases that are partitioned which gets added on reduce</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u5206\u5272\u3055\u308c\u305f\u8ca0\u8377\u30d0\u30a4\u30a2\u30b9\u304c\u30ea\u30c7\u30e5\u30fc\u30b9\u6642\u306b\u8ffd\u52a0\u3055\u308c\u308b</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u540d\u524d\u3067\u3059</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c 1 \u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li>\n<li><span translate=no>_^_3_^_</span>2 \u756a\u76ee\u306e\u30d1\u30fc\u30c6\u30a3\u30b7\u30e7\u30f3\u8f9e\u66f8</li></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Download </p>\n": "<p>[\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9]</p>\n",
|
||||
"<p>Download path </p>\n": "<p>\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u30d1\u30b9</p>\n",
|
||||
"<p>Embedding layer </p>\n": "<p>\u57cb\u3081\u8fbc\u307f\u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>Empty states (not used) </p>\n": "<p>\u7a7a\u306e\u72b6\u614b (\u672a\u4f7f\u7528)</p>\n",
|
||||
"<p>Final normalization layer and readout layer </p>\n": "<p>\u6700\u7d42\u6b63\u898f\u5316\u5c64\u3068\u8aad\u307f\u51fa\u3057\u5c64</p>\n",
|
||||
"<p>Get files to download </p>\n": "<p>\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3059\u308b\u30d5\u30a1\u30a4\u30eb\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Iterate </p>\n": "<p>\u7e70\u308a\u8fd4\u3057</p>\n",
|
||||
"<p>Layer checkpoints </p>\n": "<p>\u30ec\u30a4\u30e4\u30fc\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8</p>\n",
|
||||
"<p>Log </p>\n": "<p>\u30ed\u30b0</p>\n",
|
||||
"<p>Parent url </p>\n": "<p>\u4fdd\u8b77\u8005\u306e URL</p>\n",
|
||||
"<p>Transformer layers </p>\n": "<p>\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u5c64</p>\n",
|
||||
"<p>Vocabulary and configs </p>\n": "<p>\u8a9e\u5f59\u3068\u69cb\u6210</p>\n",
|
||||
"Code to download checkpoints and helpers to load them.": "\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3068\u305d\u308c\u3089\u3092\u30ed\u30fc\u30c9\u3059\u308b\u30d8\u30eb\u30d1\u30fc\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3059\u308b\u30b3\u30fc\u30c9\u3002",
|
||||
"GPT-NeoX Checkpoints": "GPT-\u30cd\u30aa\u30c3\u30af\u30b9\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8"
|
||||
}
|
||||
@@ -0,0 +1,25 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Checkpoints</h1>\n": "<h1>\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca\u0db8\u0dd4\u0dbb\u0db4\u0ddc\u0dbd\u0dc0\u0dbd\u0dca</h1>\n",
|
||||
"<h2>Download all checkpoint files</h2>\n": "<h2>\u0dc3\u0dd2\u0dba\u0dbd\u0dd4\u0db8\u0db4\u0dd2\u0dbb\u0dd2\u0d9a\u0dca\u0dc3\u0dd4\u0db8\u0dca \u0d9c\u0ddc\u0db1\u0dd4 \u0db6\u0dcf\u0d9c\u0db1\u0dca\u0db1</h2>\n",
|
||||
"<h3>Get files to download</h3>\n<ul><p><em>Returns</em> a list of files to be downloaded</p></ul>\n": "<h3>\u0db6\u0dcf\u0d9c\u0dad\u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0d9c\u0ddc\u0db1\u0dd4 \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1</h3>\n<ul><p>\u0db6\u0dcf\u0d9c\u0dad\u0d9a\u0dc5 \u0dba\u0dd4\u0dad\u0dd4 \u0d9c\u0ddc\u0db1\u0dd4 \u0dbd\u0dd0\u0dba\u0dd2\u0dc3\u0dca\u0dad\u0dd4\u0dc0\u0d9a\u0dca<em>\u0d86\u0db4\u0dc3\u0dd4</em> \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</p></ul>\n",
|
||||
"<h3>Load a pair of checkpoint files</h3>\n<ul><li><span translate=no>_^_0_^_</span> pair of files to load </li>\n<p><em>Returns</em> the loaded parameter tensors</p></ul>\n": "<h3>\u0db8\u0dd4\u0dbb\u0db4\u0ddc\u0dbd\u0dc0\u0dbd\u0dca\u0d9c\u0ddc\u0db1\u0dd4 \u0dba\u0dd4\u0d9c\u0dbd\u0dba\u0d9a\u0dca \u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0dd0\u0da7\u0dc0\u0dd3\u0db8\u0da7 \u0d9c\u0ddc\u0db1\u0dd4 \u0dba\u0dd4\u0d9c\u0dbd </li>\n<p>\u0db4\u0da7\u0dc0\u0db1\u0dbd\u0daf \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2 \u0d86\u0dad\u0dad\u0dd3\u0db1\u0dca<em>\u0db1\u0dd0\u0dc0\u0dad \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<h3>Load a parameter by merging the partitions along first dimension</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u0db4\u0dc5\u0db8\u0dd4\u0db8\u0dcf\u0db1\u0dba \u0d94\u0dc3\u0dca\u0dc3\u0dda \u0d9a\u0ddc\u0da7\u0dc3\u0dca \u0d92\u0d9a\u0dcf\u0db6\u0daf\u0dca\u0db0 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dd9\u0db1\u0dca \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba\u0d9a\u0dca \u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0db1\u0db8 \u0dc0\u0dda </li>\n<li><span translate=no>_^_2_^_</span> \u0db4\u0dc5\u0db8\u0dd4 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba </li>\n<li><span translate=no>_^_3_^_</span> \u0daf\u0dd9\u0dc0\u0db1 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba</li></ul>\n",
|
||||
"<h3>Load a parameter by merging the partitions along second dimension</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u0daf\u0dd9\u0dc0\u0db1\u0db8\u0dcf\u0db1\u0dba \u0d94\u0dc3\u0dca\u0dc3\u0dda \u0d9a\u0ddc\u0da7\u0dc3\u0dca \u0d92\u0d9a\u0dcf\u0db6\u0daf\u0dca\u0db0 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dd9\u0db1\u0dca \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba\u0d9a\u0dca \u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0db1\u0db8 \u0dc0\u0dda </li>\n<li><span translate=no>_^_2_^_</span> \u0db4\u0dc5\u0db8\u0dd4 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba </li>\n<li><span translate=no>_^_3_^_</span> \u0daf\u0dd9\u0dc0\u0db1 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba</li></ul>\n",
|
||||
"<h3>Load an un-partitioned parameter</h3>\n<p>This does a sanity check to make use both partitions are the same</p>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u0d9a\u0d91\u0d9a\u0dca\u0dc3\u0dad\u0dca -\u0d9a\u0ddc\u0da7\u0dc3\u0dca \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0db4\u0dd6\u0dbb\u0dab\u0dba</h3>\n<p>\u0d9a\u0ddc\u0da7\u0dc3\u0dca\u0daf\u0dd9\u0d9a\u0db8 \u0d91\u0d9a \u0dc4\u0dcf \u0dc3\u0db8\u0dcf\u0db1 \u0dc0\u0db1 \u0db4\u0dbb\u0dd2\u0daf\u0dd2 \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0db8\u0dd9\u0dba \u0dc3\u0db1\u0dd3\u0db4\u0dcf\u0dbb\u0d9a\u0dca\u0dc2\u0d9a \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dab\u0dba\u0d9a\u0dca \u0dc3\u0dd2\u0daf\u0dd4 \u0d9a\u0dbb\u0dba\u0dd2</p>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0db1\u0db8 \u0dc0\u0dda </li>\n<li><span translate=no>_^_2_^_</span> \u0db4\u0dc5\u0db8\u0dd4 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba </li>\n<li><span translate=no>_^_3_^_</span> \u0daf\u0dd9\u0dc0\u0db1 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba</li></ul>\n",
|
||||
"<h3>Load biases that are partitioned which gets added on reduce</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u0d9a\u0ddc\u0da7\u0dc3\u0dca\u0d9a\u0dbb \u0d87\u0dad\u0dd2 \u0db6\u0dc0 \u0d85\u0d9c\u0dad\u0dd3\u0db1\u0dca \u0db4\u0dd0\u0da7\u0dc0\u0dd3\u0db8 \u0d85\u0da9\u0dd4 \u0db8\u0dad \u0d91\u0d9a\u0dad\u0dd4 \u0dbd\u0dd0\u0db6\u0dd9\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2\u0dba \u0db1\u0db8 \u0dc0\u0dda </li>\n<li><span translate=no>_^_2_^_</span> \u0db4\u0dc5\u0db8\u0dd4 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba </li>\n<li><span translate=no>_^_3_^_</span> \u0daf\u0dd9\u0dc0\u0db1 \u0d9a\u0ddc\u0da7\u0dc3 \u0dc1\u0db6\u0dca\u0daf \u0d9a\u0ddd\u0dc2\u0dba</li></ul>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Download </p>\n": "<p>\u0db6\u0dcf\u0d9c\u0dad </p>\n",
|
||||
"<p>Download path </p>\n": "<p>\u0db6\u0dcf\u0d9c\u0dad\u0db8\u0dcf\u0dbb\u0dca\u0d9c\u0dba </p>\n",
|
||||
"<p>Embedding layer </p>\n": "<p>\u0d9a\u0dcf\u0dc0\u0dd0\u0daf\u0dca\u0daf\u0dd3\u0db8\u0dc3\u0dca\u0dae\u0dbb\u0dba </p>\n",
|
||||
"<p>Empty states (not used) </p>\n": "<p>\u0dc4\u0dd2\u0dc3\u0dca\u0dad\u0dad\u0dca\u0dc0\u0dba\u0db1\u0dca (\u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0db1\u0ddc\u0d9a\u0dd9\u0dbb\u0dda) </p>\n",
|
||||
"<p>Final normalization layer and readout layer </p>\n": "<p>\u0d85\u0dc0\u0dc3\u0dcf\u0db1\u0dc3\u0dcf\u0db8\u0dcf\u0db1\u0dca\u0dba\u0d9a\u0dbb\u0dab \u0dc3\u0dca\u0dad\u0dbb\u0dba \u0dc3\u0dc4 \u0d9a\u0dd2\u0dba\u0dc0\u0dd3\u0db8\u0dda \u0dc3\u0dca\u0dae\u0dbb\u0dba </p>\n",
|
||||
"<p>Get files to download </p>\n": "<p>\u0db6\u0dcf\u0d9c\u0dad\u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0d9c\u0ddc\u0db1\u0dd4 \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Iterate </p>\n": "<p>\u0db4\u0dd4\u0db1\u0dbb\u0dcf\u0dc0\u0dbb\u0dca\u0dad\u0db1\u0dba\u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Layer checkpoints </p>\n": "<p>\u0dc3\u0dca\u0dae\u0dbb\u0db8\u0dd4\u0dbb\u0db4\u0ddc\u0dbd\u0dc0\u0dbd\u0dca </p>\n",
|
||||
"<p>Log </p>\n": "<p>\u0dbd\u0ddc\u0d9c\u0dca </p>\n",
|
||||
"<p>Parent url </p>\n": "<p>\u0daf\u0dd9\u0db8\u0dcf\u0db4\u0dd2\u0dbaurl </p>\n",
|
||||
"<p>Transformer layers </p>\n": "<p>\u0da7\u0dca\u0dbb\u0dcf\u0db1\u0dca\u0dc3\u0dca\u0dc6\u0ddd\u0db8\u0dbb\u0dca\u0dc3\u0dca\u0dae\u0dbb </p>\n",
|
||||
"<p>Vocabulary and configs </p>\n": "<p>\u0dc0\u0da0\u0db1\u0db8\u0dcf\u0dbd\u0dcf\u0dc0 \u0dc3\u0dc4 \u0dc0\u0dd2\u0db1\u0dca\u0dba\u0dcf\u0dc3 </p>\n",
|
||||
"Code to download checkpoints and helpers to load them.": "\u0d92\u0dc0\u0dcf \u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0db8\u0dd4\u0dbb\u0db4\u0ddc\u0dbd\u0dc0\u0dbd\u0dca \u0dc3\u0dc4 \u0d8b\u0daf\u0dc0\u0dca\u0d9a\u0dbb\u0dd4\u0dc0\u0db1\u0dca \u0db6\u0dcf\u0d9c\u0dad \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0d9a\u0dda\u0dad\u0dba.",
|
||||
"GPT-NeoX Checkpoints": "\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0db8\u0dd4\u0dbb\u0db4\u0ddc\u0dbd\u0dc0\u0dbd\u0dca"
|
||||
}
|
||||
@@ -0,0 +1,25 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Checkpoints</h1>\n": "<h1>GPT-neox \u68c0\u67e5\u70b9</h1>\n",
|
||||
"<h2>Download all checkpoint files</h2>\n": "<h2>\u4e0b\u8f7d\u6240\u6709\u68c0\u67e5\u70b9\u6587\u4ef6</h2>\n",
|
||||
"<h3>Get files to download</h3>\n<ul><p><em>Returns</em> a list of files to be downloaded</p></ul>\n": "<h3>\u83b7\u53d6\u8981\u4e0b\u8f7d\u7684\u6587\u4ef6</h3>\n<ul><p><em>\u8fd4\u56de</em>\u8981\u4e0b\u8f7d\u7684\u6587\u4ef6\u5217\u8868</p></ul>\n",
|
||||
"<h3>Load a pair of checkpoint files</h3>\n<ul><li><span translate=no>_^_0_^_</span> pair of files to load </li>\n<p><em>Returns</em> the loaded parameter tensors</p></ul>\n": "<h3>\u52a0\u8f7d\u4e00\u5bf9\u68c0\u67e5\u70b9\u6587\u4ef6</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u4e00\u5bf9\u8981\u52a0\u8f7d\u7684\u6587\u4ef6</li>\n<p><em>\u8fd4\u56de</em>\u52a0\u8f7d\u7684\u53c2\u6570\u5f20\u91cf</p></ul>\n",
|
||||
"<h3>Load a parameter by merging the partitions along first dimension</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u901a\u8fc7\u5408\u5e76\u6cbf\u7b2c\u4e00\u7ef4\u5ea6\u7684\u5206\u533a\u6765\u52a0\u8f7d\u53c2\u6570</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u53c2\u6570</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u53c2\u6570\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c\u4e00\u4e2a\u5206\u533a\u5b57\u5178</li>\n<li><span translate=no>_^_3_^_</span>\u7b2c\u4e8c\u4e2a\u5206\u533a\u5b57\u5178</li></ul>\n",
|
||||
"<h3>Load a parameter by merging the partitions along second dimension</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u901a\u8fc7\u5408\u5e76\u7b2c\u4e8c\u7ef4\u5ea6\u7684\u5206\u533a\u6765\u52a0\u8f7d\u53c2\u6570</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u53c2\u6570</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u53c2\u6570\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c\u4e00\u4e2a\u5206\u533a\u5b57\u5178</li>\n<li><span translate=no>_^_3_^_</span>\u7b2c\u4e8c\u4e2a\u5206\u533a\u5b57\u5178</li></ul>\n",
|
||||
"<h3>Load an un-partitioned parameter</h3>\n<p>This does a sanity check to make use both partitions are the same</p>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u52a0\u8f7d\u672a\u5206\u533a\u7684\u53c2\u6570</h3>\n<p>\u8fd9\u4f1a\u8fdb\u884c\u5065\u5168\u6027\u68c0\u67e5\uff0c\u4ee5\u4f7f\u7528\u4e24\u4e2a\u5206\u533a\u662f\u76f8\u540c\u7684</p>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u53c2\u6570</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u53c2\u6570\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c\u4e00\u4e2a\u5206\u533a\u5b57\u5178</li>\n<li><span translate=no>_^_3_^_</span>\u7b2c\u4e8c\u4e2a\u5206\u533a\u5b57\u5178</li></ul>\n",
|
||||
"<h3>Load biases that are partitioned which gets added on reduce</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the parameter </li>\n<li><span translate=no>_^_1_^_</span> is the name of the parameter </li>\n<li><span translate=no>_^_2_^_</span> first partition dictionary </li>\n<li><span translate=no>_^_3_^_</span> second partition dictionary</li></ul>\n": "<h3>\u5206\u533a\u7684\u8d1f\u8f7d\u504f\u5dee\u5728 reduce \u65f6\u88ab\u6dfb\u52a0</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u53c2\u6570</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u53c2\u6570\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_2_^_</span>\u7b2c\u4e00\u4e2a\u5206\u533a\u5b57\u5178</li>\n<li><span translate=no>_^_3_^_</span>\u7b2c\u4e8c\u4e2a\u5206\u533a\u5b57\u5178</li></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Download </p>\n": "<p>\u4e0b\u8f7d</p>\n",
|
||||
"<p>Download path </p>\n": "<p>\u4e0b\u8f7d\u8def\u5f84</p>\n",
|
||||
"<p>Embedding layer </p>\n": "<p>\u5d4c\u5165\u5c42</p>\n",
|
||||
"<p>Empty states (not used) </p>\n": "<p>\u7a7a\u72b6\u6001\uff08\u672a\u4f7f\u7528\uff09</p>\n",
|
||||
"<p>Final normalization layer and readout layer </p>\n": "<p>\u6700\u7ec8\u5f52\u4e00\u5316\u5c42\u548c\u8bfb\u51fa\u5c42</p>\n",
|
||||
"<p>Get files to download </p>\n": "<p>\u83b7\u53d6\u8981\u4e0b\u8f7d\u7684\u6587\u4ef6</p>\n",
|
||||
"<p>Iterate </p>\n": "<p>\u8fed\u4ee3</p>\n",
|
||||
"<p>Layer checkpoints </p>\n": "<p>\u56fe\u5c42\u68c0\u67e5\u70b9</p>\n",
|
||||
"<p>Log </p>\n": "<p>\u65e5\u5fd7</p>\n",
|
||||
"<p>Parent url </p>\n": "<p>\u5bb6\u957f\u7f51\u5740</p>\n",
|
||||
"<p>Transformer layers </p>\n": "<p>\u53d8\u538b\u5668\u5c42</p>\n",
|
||||
"<p>Vocabulary and configs </p>\n": "<p>\u8bcd\u6c47\u548c\u914d\u7f6e</p>\n",
|
||||
"Code to download checkpoints and helpers to load them.": "\u4e0b\u8f7d\u68c0\u67e5\u70b9\u7684\u4ee3\u7801\u548c\u52a0\u8f7d\u5b83\u4eec\u7684\u52a9\u624b\u3002",
|
||||
"GPT-NeoX Checkpoints": "GPT-neox \u68c0\u67e5\u70b9"
|
||||
}
|
||||
@@ -0,0 +1,54 @@
|
||||
{
|
||||
"<h1>Evaluation</h1>\n<p>This is the code to test the model on <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">EleutherAI/lm-evaluation-harness</a>.</p>\n<ul><li><a href=\"half_precision.html\">Evaluating half precision model on a single GPU</a></li></ul>\n": "<h1>\u8a55\u4fa1</h1>\n<p><a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">Eleutherai/LM-\u8a55\u4fa1\u30cf\u30fc\u30cd\u30b9\u3067\u30e2\u30c7\u30eb\u3092\u30c6\u30b9\u30c8\u3059\u308b\u305f\u3081\u306e\u30b3\u30fc\u30c9\u3067\u3059</a>\u3002</p>\n<ul><li><a href=\"half_precision.html\">\u5358\u4e00\u306e GPU \u3067\u306e\u534a\u7cbe\u5ea6\u30e2\u30c7\u30eb\u306e\u8a55\u4fa1</a></li></ul>\n",
|
||||
"<h2>Evaluation Harness Adapter</h2>\n<p>This is based on the <a href=\"https://github.com/EleutherAI/gpt-neox/blob/main/eval_tasks/eval_adapter.py\">adapter from EleutherAI/gpt-neox</a></p>\n": "<h2>\u8a55\u4fa1\u7528\u30cf\u30fc\u30cd\u30b9\u30a2\u30c0\u30d7\u30bf\u30fc</h2>\n<p><a href=\"https://github.com/EleutherAI/gpt-neox/blob/main/eval_tasks/eval_adapter.py\">\u3053\u308c\u306fElutherai/GPT-Neox\u306e\u30a2\u30c0\u30d7\u30bf\u30fc\u3092\u30d9\u30fc\u30b9\u306b\u3057\u3066\u3044\u307e\u3059</a></p>\n",
|
||||
"<h2>Run evaluation harness with a given model</h2>\n": "<h2>\u7279\u5b9a\u306e\u30e2\u30c7\u30eb\u3067\u8a55\u4fa1\u7528\u30cf\u30fc\u30cd\u30b9\u3092\u5b9f\u884c</h2>\n",
|
||||
"<h3>Get log-likelihoods of the next tokens</h3>\n<ul><li><span translate=no>_^_0_^_</span> List of requests containing the context and the expected continuation. </li>\n<li><span translate=no>_^_1_^_</span> If True, disable tqdm progress bar.</li></ul>\n": "<h3>\u6b21\u306e\u30c8\u30fc\u30af\u30f3\u306e\u5bfe\u6570\u767a\u751f\u78ba\u7387\u3092\u53d6\u5f97</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u3068\u4e88\u60f3\u3055\u308c\u308b\u7d99\u7d9a\u3092\u542b\u3080\u30ea\u30af\u30a8\u30b9\u30c8\u306e\u30ea\u30b9\u30c8\u3002</li>\n</ul><li><span translate=no>_^_1_^_</span>True \u306e\u5834\u5408\u3001tqdm \u30d7\u30ed\u30b0\u30ec\u30b9\u30d0\u30fc\u3092\u7121\u52b9\u306b\u3057\u307e\u3059\u3002</li>\n",
|
||||
"<h3>Run given evaluations</h3>\n": "<h3>\u4e0e\u3048\u3089\u308c\u305f\u8a55\u4fa1\u3092\u5b9f\u884c</h3>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p> Batch size</p>\n": "<p>\u30d0\u30c3\u30c1\u30b5\u30a4\u30ba</p>\n",
|
||||
"<p> Call the model</p>\n": "<p>\u30e2\u30c7\u30eb\u306b\u96fb\u8a71\u3059\u308b</p>\n",
|
||||
"<p> Decode text from token ids</p>\n": "<p>\u30c8\u30fc\u30af\u30f3 ID \u304b\u3089\u30c6\u30ad\u30b9\u30c8\u3092\u30c7\u30b3\u30fc\u30c9</p>\n",
|
||||
"<p> Encode a given text</p>\n": "<p>\u4e0e\u3048\u3089\u308c\u305f\u30c6\u30ad\u30b9\u30c8\u3092\u30a8\u30f3\u30b3\u30fc\u30c9\u3059\u308b</p>\n",
|
||||
"<p>Add configs </p>\n": "<p>\u69cb\u6210\u3092\u8ffd\u52a0</p>\n",
|
||||
"<p>Add padding </p>\n": "<p>\u30d1\u30c7\u30a3\u30f3\u30b0\u3092\u8ffd\u52a0</p>\n",
|
||||
"<p>Add the total log-likelihoods and whether there was a match to the results </p>\n": "<p>\u5bfe\u6570\u63a8\u5b9a\u5024\u306e\u5408\u8a08\u3068\u3001\u4e00\u81f4\u3057\u305f\u304b\u3069\u3046\u304b\u3092\u7d50\u679c\u306b\u52a0\u7b97\u3057\u307e\u3059\u3002</p>\n",
|
||||
"<p>All tasks if nothing is specified </p>\n": "<p>\u4f55\u3082\u6307\u5b9a\u3055\u308c\u3066\u3044\u306a\u3044\u5834\u5408\u306f\u3059\u3079\u3066\u306e\u30bf\u30b9\u30af</p>\n",
|
||||
"<p>Concatenate the context and continuation </p>\n": "<p>\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u3068\u7d9a\u304d\u3092\u9023\u7d50\u3059\u308b</p>\n",
|
||||
"<p>Create a tensor </p>\n": "<p>\u30c6\u30f3\u30bd\u30eb\u306e\u4f5c\u6210</p>\n",
|
||||
"<p>Create the adapter </p>\n": "<p>\u30a2\u30c0\u30d7\u30bf\u30fc\u306e\u4f5c\u6210</p>\n",
|
||||
"<p>Determine the padded length. Shorter sequences will get padded. </p>\n": "<p>\u30d1\u30c3\u30c9\u306e\u9577\u3055\u3092\u6c7a\u3081\u3066\u304f\u3060\u3055\u3044\u3002\u77ed\u3044\u30b7\u30fc\u30b1\u30f3\u30b9\u306f\u30d1\u30c7\u30a3\u30f3\u30b0\u3055\u308c\u307e\u3059</p>\u3002\n",
|
||||
"<p>End-of-text token </p>\n": "<p>\u30c6\u30ad\u30b9\u30c8\u7d42\u4e86\u30c8\u30fc\u30af\u30f3</p>\n",
|
||||
"<p>For results </p>\n": "<p>\u7d50\u679c\u306b\u3064\u3044\u3066</p>\n",
|
||||
"<p>Get log softmaxes </p>\n": "<p>\u30ed\u30b0\u30bd\u30d5\u30c8\u30de\u30c3\u30af\u30b9\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get logits of those </p>\n": "<p>\u305d\u308c\u3089\u306e\u30ed\u30b0\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get model logits </p>\n": "<p>\u30e2\u30c7\u30eb\u30ed\u30b8\u30c3\u30c8\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get number of predicted tokens </p>\n": "<p>\u4e88\u6e2c\u30c8\u30fc\u30af\u30f3\u306e\u6570\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get the target tokens </p>\n": "<p>\u5bfe\u8c61\u30c8\u30fc\u30af\u30f3\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get the tokens with the highest probabilities </p>\n": "<p>\u6700\u3082\u78ba\u7387\u306e\u9ad8\u3044\u30c8\u30fc\u30af\u30f3\u3092\u624b\u306b\u5165\u308c\u3088\u3046</p>\n",
|
||||
"<p>Input length </p>\n": "<p>\u5165\u529b\u9577\u3055</p>\n",
|
||||
"<p>Lengths of the input sequences </p>\n": "<p>\u5165\u529b\u30b7\u30fc\u30b1\u30f3\u30b9\u306e\u9577\u3055</p>\n",
|
||||
"<p>Load the tokenizer </p>\n": "<p>\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u30ed\u30fc\u30c9</p>\n",
|
||||
"<p>Log-likelihoods of the target tokens </p>\n": "<p>\u5bfe\u8c61\u30c8\u30fc\u30af\u30f3\u306e\u5bfe\u6570\u767a\u751f\u53ef\u80fd\u6027</p>\n",
|
||||
"<p>Loop through each request in the chunk and collect them into PyTorch tensors with paddings </p>\n": "<p>\u30c1\u30e3\u30f3\u30af\u5185\u306e\u5404\u30ea\u30af\u30a8\u30b9\u30c8\u3092\u30eb\u30fc\u30d7\u51e6\u7406\u3057\u3001\u30d1\u30c7\u30a3\u30f3\u30b0\u4ed8\u304d\u306e PyTorch \u30c6\u30f3\u30bd\u30eb\u306b\u307e\u3068\u3081\u307e\u3059\u3002</p>\n",
|
||||
"<p>Loop through requests with <span translate=no>_^_0_^_</span> number of requests at a time </p>\n": "<p><span translate=no>_^_0_^_</span>\u4e00\u5ea6\u306b\u8907\u6570\u306e\u30ea\u30af\u30a8\u30b9\u30c8\u304c\u3042\u308b\u30ea\u30af\u30a8\u30b9\u30c8\u3092\u30eb\u30fc\u30d7\u30b9\u30eb\u30fc\u3059\u308b</p>\n",
|
||||
"<p>Loop through the input/output pairs of the batch </p>\n": "<p>\u30d0\u30c3\u30c1\u306e\u5165\u529b\u3068\u51fa\u529b\u306e\u30da\u30a2\u3092\u30eb\u30fc\u30d7\u51e6\u7406\u3057\u307e\u3059</p>\n",
|
||||
"<p>Maximum number of tokens to generate </p>\n": "<p>\u751f\u6210\u3059\u308b\u30c8\u30fc\u30af\u30f3\u306e\u6700\u5927\u6570</p>\n",
|
||||
"<p>Maximum sequence length </p>\n": "<p>\u6700\u5927\u30b7\u30fc\u30b1\u30f3\u30b9\u9577</p>\n",
|
||||
"<p>Padded length for the batch </p>\n": "<p>\u30d0\u30c3\u30c1\u7528\u306e\u30d1\u30c3\u30c9\u5165\u308a\u9577\u3055</p>\n",
|
||||
"<p>Padding </p>\n": "<p>\u30d1\u30c7\u30a3\u30f3\u30b0</p>\n",
|
||||
"<p>Re-order and return results </p>\n": "<p>\u4e26\u3079\u66ff\u3048\u3066\u7d50\u679c\u3092\u8fd4\u3059</p>\n",
|
||||
"<p>Remove final token </p>\n": "<p>\u6700\u7d42\u30c8\u30fc\u30af\u30f3\u3092\u524a\u9664</p>\n",
|
||||
"<p>Reorder the requests in the descending order of the lengths, so that sequences with similar lengths are close </p>\n": "<p>\u540c\u3058\u9577\u3055\u306e\u30b7\u30fc\u30b1\u30f3\u30b9\u304c\u8fd1\u304f\u306a\u308b\u3088\u3046\u306b\u3001\u30ea\u30af\u30a8\u30b9\u30c8\u3092\u9577\u3055\u306e\u964d\u9806\u306b\u4e26\u3079\u66ff\u3048\u307e\u3059</p>\n",
|
||||
"<p>Run </p>\n": "<p>\u5b9f\u884c</p>\n",
|
||||
"<p>Run <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">EleutherAI/lm-evaluation-harness</a> evaluator </p>\n": "<p><a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">EleutherAI/LM-\u8a55\u4fa1-\u30cf\u30fc\u30cd\u30b9\u30a8\u30d0\u30ea\u30e5\u30a8\u30fc\u30bf\u30fc\u3092\u5b9f\u884c</a></p>\n",
|
||||
"<p>Size of the vocabulary </p>\n": "<p>\u30dc\u30ad\u30e3\u30d6\u30e9\u30ea\u30fc\u306e\u30b5\u30a4\u30ba</p>\n",
|
||||
"<p>The continuations for the batch </p>\n": "<p>\u30d0\u30c3\u30c1\u306e\u7d99\u7d9a</p>\n",
|
||||
"<p>To store the inputs for the batch </p>\n": "<p>\u30d0\u30c3\u30c1\u306e\u5165\u529b\u3092\u4fdd\u5b58\u3059\u308b\u306b\u306f</p>\n",
|
||||
"<p>Truncate from left if the size exceeds the <span translate=no>_^_0_^_</span> </p>\n": "<p>\u30b5\u30a4\u30ba\u304c <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Whether there's an exact match </p>\n": "<p>\u5b8c\u5168\u306b\u4e00\u81f4\u3059\u308b\u304b\u3069\u3046\u304b</p>\n",
|
||||
"<p>padded_length = padded_length if padded_length is not None else inplen </p>\n": "<p>padded_length = padded_length \u304c Padded_length \u3067\u306a\u3044\u5834\u5408\u306f\u30d1\u30c7\u30a3\u30f3\u30b0\u3055\u308c\u305f_length\u3001\u305d\u308c\u4ee5\u5916\u306f\u30d7\u30ec\u30f3\u306a\u3057</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is model </li>\n<li><span translate=no>_^_1_^_</span> is the <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a> </li>\n<li><span translate=no>_^_2_^_</span> is the size of the vocabulary (this differs from the tokenizer vocab size since neox adds some extra to make the embedding layer model parallel.) </li>\n<li><span translate=no>_^_3_^_</span> is the batch size </li>\n<li><span translate=no>_^_4_^_</span> is the device of the model</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u30e2\u30c7\u30eb\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span><a href=\"huggingface/tokenizers\">\u30cf\u30ae\u30f3\u30b0\u30d5\u30a7\u30a4\u30b9\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3067\u3059</a></li>\n<li><span translate=no>_^_2_^_</span>\u306f\u30dc\u30ad\u30e3\u30d6\u30e9\u30ea\u306e\u30b5\u30a4\u30ba\u3067\u3059 (\u3053\u308c\u306f\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u306e\u30dc\u30ad\u30e3\u30d6\u30b5\u30a4\u30ba\u3068\u306f\u7570\u306a\u308a\u307e\u3059\u3002neox\u306f\u57cb\u3081\u8fbc\u307f\u5c64\u30e2\u30c7\u30eb\u3092\u4e26\u5217\u5316\u3059\u308b\u305f\u3081\u306e\u8ffd\u52a0\u6a5f\u80fd\u3092\u8ffd\u52a0\u3057\u3066\u3044\u308b\u304b\u3089\u3067\u3059)\u3002</li>\n<li><span translate=no>_^_3_^_</span>\u306f\u30d0\u30c3\u30c1\u30b5\u30a4\u30ba</li>\n<li><span translate=no>_^_4_^_</span>\u30e2\u30c7\u30eb\u306e\u30c7\u30d0\u30a4\u30b9\u3067\u3059</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a> </li>\n<li><span translate=no>_^_1_^_</span> is the size of the vocabulary (this differs from the tokenizer vocab size since neox adds some extra to make the embedding layer model parallel.) </li>\n<li><span translate=no>_^_2_^_</span> is the batch size</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span><a href=\"huggingface/tokenizers\">\u30cf\u30ae\u30f3\u30b0\u30d5\u30a7\u30a4\u30b9\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3067\u3059</a></li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30dc\u30ad\u30e3\u30d6\u30e9\u30ea\u306e\u30b5\u30a4\u30ba\u3067\u3059 (\u3053\u308c\u306f\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u306e\u30dc\u30ad\u30e3\u30d6\u30b5\u30a4\u30ba\u3068\u306f\u7570\u306a\u308a\u307e\u3059\u3002neox\u306f\u57cb\u3081\u8fbc\u307f\u5c64\u30e2\u30c7\u30eb\u3092\u4e26\u5217\u5316\u3059\u308b\u305f\u3081\u306e\u8ffd\u52a0\u6a5f\u80fd\u3092\u8ffd\u52a0\u3057\u3066\u3044\u308b\u304b\u3089\u3067\u3059)\u3002</li>\n<li><span translate=no>_^_2_^_</span>\u306f\u30d0\u30c3\u30c1\u30b5\u30a4\u30ba</li></ul>\n",
|
||||
"Code to evaluate the model on NLP tasks through lm-evaluation-harness": "LM \u8a55\u4fa1\u30cf\u30fc\u30cd\u30b9\u3092\u901a\u3058\u3066 NLP \u30bf\u30b9\u30af\u3067\u30e2\u30c7\u30eb\u3092\u8a55\u4fa1\u3059\u308b\u30b3\u30fc\u30c9",
|
||||
"Evaluation": "\u8a55\u4fa1"
|
||||
}
|
||||
@@ -0,0 +1,54 @@
|
||||
{
|
||||
"<h1>Evaluation</h1>\n<p>This is the code to test the model on <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">EleutherAI/lm-evaluation-harness</a>.</p>\n<ul><li><a href=\"half_precision.html\">Evaluating half precision model on a single GPU</a></li></ul>\n": "<h1>\u0d87\u0d9c\u0dba\u0dd3\u0db8</h1>\n<p><a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">Eleutherai/LM-\u0d87\u0d9c\u0dba\u0dd3\u0db8\u0dca-\u0db4\u0da7\u0dd2</a>\u0db4\u0dd2\u0dc5\u0dd2\u0db6\u0db3 \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda \u0d9a\u0dda\u0dad\u0dba \u0db8\u0dd9\u0dba\u0dba\u0dd2. </p>\n<ul><li><a href=\"half_precision.html\">\u0dad\u0db1\u0dd2 GPU \u0db8\u0dad \u0d85\u0dbb\u0dca\u0db0 \u0db1\u0dd2\u0dbb\u0dc0\u0daf\u0dca\u0dba\u0dad\u0dcf \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0d9a\u0dca \u0d87\u0d9c\u0dba\u0dd3\u0db8</a></li></ul>\n",
|
||||
"<h2>Evaluation Harness Adapter</h2>\n<p>This is based on the <a href=\"https://github.com/EleutherAI/gpt-neox/blob/main/eval_tasks/eval_adapter.py\">adapter from EleutherAI/gpt-neox</a></p>\n": "<h2>\u0d87\u0d9c\u0dba\u0dd3\u0db8\u0dca\u0db4\u0da7\u0dd2 \u0d87\u0da9\u0db4\u0dca\u0da7\u0dbb</h2>\n<p>\u0db8\u0dd9\u0dba <a href=\"https://github.com/EleutherAI/gpt-neox/blob/main/eval_tasks/eval_adapter.py\">Eleutherai/GPT-neox \u0dc0\u0dd9\u0dad\u0dd2\u0db1\u0dca \u0d87\u0da9\u0db4\u0dca\u0da7\u0dbb\u0dba</a>\u0db8\u0dad \u0db4\u0daf\u0db1\u0db8\u0dca \u0dc0\u0dda</p>\n",
|
||||
"<h2>Run evaluation harness with a given model</h2>\n": "<h2>\u0daf\u0dd3\u0d87\u0dad\u0dd2 \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0d9a\u0dca \u0dc3\u0db8\u0d9f \u0d87\u0d9c\u0dba\u0dd3\u0db8\u0dca \u0db4\u0da7\u0dd2 \u0db0\u0dcf\u0dc0\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h2>\n",
|
||||
"<h3>Get log-likelihoods of the next tokens</h3>\n<ul><li><span translate=no>_^_0_^_</span> List of requests containing the context and the expected continuation. </li>\n<li><span translate=no>_^_1_^_</span> If True, disable tqdm progress bar.</li></ul>\n": "<h3>\u0d8a\u0dc5\u0d9f\u0da7\u0ddd\u0d9a\u0db1 \u0dc0\u0dbd \u0dbd\u0ddc\u0d9c\u0dca \u0dc0\u0dd3\u0db8\u0dda \u0dc3\u0db8\u0dcf\u0db1\u0d9a\u0db8\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0dc3\u0db1\u0dca\u0daf\u0dbb\u0dca\u0db7\u0dba \u0dc3\u0dc4 \u0d85\u0db4\u0dda\u0d9a\u0dca\u0dc2\u0dd2\u0dad \u0d85\u0d9b\u0dab\u0dca\u0da9 \u0db4\u0dd0\u0dc0\u0dd0\u0dad\u0dca\u0db8 \u0d85\u0da9\u0d82\u0d9c\u0dd4 \u0d89\u0dbd\u0dca\u0dbd\u0dd3\u0db8\u0dca \u0dbd\u0dd0\u0dba\u0dd2\u0dc3\u0dca\u0dad\u0dd4\u0dc0. </li>\n<li><span translate=no>_^_1_^_</span> \u0dc3\u0dad\u0dca\u0dba \u0db1\u0db8\u0dca, tqdm \u0db4\u0dca\u0dbb\u0d9c\u0dad\u0dd2 \u0dad\u0dd3\u0dbb\u0dd4\u0dc0 \u0d85\u0d9a\u0dca\u0dbb\u0dd3\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1. </li></ul>\n",
|
||||
"<h3>Run given evaluations</h3>\n": "<h3>\u0dbd\u0db6\u0dcf\u0daf\u0dd3 \u0d87\u0dad\u0dd2 \u0d87\u0d9c\u0dba\u0dd3\u0db8\u0dca \u0d9a\u0dca\u0dbb\u0dd2\u0dba\u0dcf\u0dad\u0dca\u0db8\u0d9a \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p> Batch size</p>\n": "<p> \u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8\u0dca\u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba</p>\n",
|
||||
"<p> Call the model</p>\n": "<p> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0d85\u0db8\u0dad\u0db1\u0dca\u0db1</p>\n",
|
||||
"<p> Decode text from token ids</p>\n": "<p> \u0da7\u0ddd\u0d9a\u0db1\u0dca\u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca\u0db4\u0dad\u0dca \u0dc0\u0dbd\u0dd2\u0db1\u0dca \u0db4\u0dd9\u0dc5 \u0dc0\u0dd2\u0d9a\u0dda\u0dad\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</p>\n",
|
||||
"<p> Encode a given text</p>\n": "<p> \u0daf\u0dd3\u0d87\u0dad\u0dd2 \u0db4\u0dd9\u0dc5\u0d9a\u0dca \u0d9a\u0dda\u0dad\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</p>\n",
|
||||
"<p>Add configs </p>\n": "<p>\u0dc0\u0dd2\u0db1\u0dca\u0dba\u0dcf\u0dc3\u0d91\u0d9a\u0dad\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Add padding </p>\n": "<p>\u0db4\u0dd1\u0da9\u0dd2\u0db1\u0dca\u0d91\u0d9a\u0dad\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Add the total log-likelihoods and whether there was a match to the results </p>\n": "<p>\u0db8\u0dd4\u0dc5\u0dd4\u0dbd\u0ddc\u0d9c\u0dca-Likehoods \u0d91\u0d9a\u0dad\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 \u0dc3\u0dc4 \u0db4\u0dca\u0dbb\u0dad\u0dd2\u0db5\u0dbd \u0dc3\u0db3\u0dc4\u0dcf \u0dad\u0dbb\u0d9c\u0dba \u0dad\u0dd2\u0db6\u0dd4\u0dab\u0dda \u0daf \u0dba\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>All tasks if nothing is specified </p>\n": "<p>\u0d9a\u0dd2\u0dc3\u0dd2\u0dc0\u0d9a\u0dca\u0db1\u0dd2\u0dba\u0db8 \u0d9a\u0dbb \u0db1\u0ddc\u0db8\u0dd0\u0dad\u0dd2 \u0db1\u0db8\u0dca \u0dc3\u0dd2\u0dba\u0dbd\u0dd4 \u0d9a\u0dcf\u0dbb\u0dca\u0dba\u0dba\u0db1\u0dca </p>\n",
|
||||
"<p>Concatenate the context and continuation </p>\n": "<p>\u0dc3\u0db1\u0dca\u0daf\u0dbb\u0dca\u0db7\u0dba\u0dc3\u0dc4 \u0d85\u0d9b\u0dab\u0dca\u0da9 \u0db4\u0dd0\u0dc0\u0dd0\u0dad\u0dca\u0db8 \u0dc3\u0d82\u0dba\u0dd4\u0d9a\u0dca\u0dad \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Create a tensor </p>\n": "<p>\u0d86\u0dad\u0dad\u0dd2\u0dba\u0d9a\u0dca\u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Create the adapter </p>\n": "<p>\u0d87\u0da9\u0dd0\u0db4\u0dca\u0da7\u0dbb\u0dba\u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Determine the padded length. Shorter sequences will get padded. </p>\n": "<p>\u0db4\u0dd1\u0da9\u0dca\u0daf\u0dd2\u0d9c \u0dad\u0dd3\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1. \u0d9a\u0dd9\u0da7\u0dd2 \u0d85\u0db1\u0dd4\u0db4\u0dd2\u0dc5\u0dd2\u0dc0\u0dd9\u0dbd\u0dc0\u0dbd\u0dca \u0db4\u0dd1\u0da9\u0dca \u0dbd\u0dd0\u0db6\u0dd9\u0db1\u0dd4 \u0d87\u0dad. </p>\n",
|
||||
"<p>End-of-text token </p>\n": "<p>\u0db4\u0dd9\u0dc5\u0d85\u0dc0\u0dc3\u0db1\u0dca \u0da7\u0ddd\u0d9a\u0db1\u0dba </p>\n",
|
||||
"<p>For results </p>\n": "<p>\u0db4\u0dca\u0dbb\u0dad\u0dd2\u0db5\u0dbd\u0dc3\u0db3\u0dc4\u0dcf </p>\n",
|
||||
"<p>Get log softmaxes </p>\n": "<p>\u0dbd\u0ddc\u0d9c\u0dca\u0dc3\u0ddc\u0dc6\u0dca\u0da7\u0dca\u0db8\u0dd0\u0d9a\u0dca\u0dc3\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Get logits of those </p>\n": "<p>\u0d85\u0dba\u0d9c\u0dda\u0db4\u0dd2\u0dc0\u0dd2\u0dc3\u0dd4\u0db8\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Get model logits </p>\n": "<p>\u0d86\u0daf\u0dbb\u0dca\u0dc1\u0db4\u0dd2\u0dc0\u0dd2\u0dc3\u0dd4\u0db8\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Get number of predicted tokens </p>\n": "<p>\u0db4\u0dd4\u0dbb\u0ddd\u0d9a\u0dae\u0db1\u0dba\u0d9a\u0dc5 \u0da7\u0ddd\u0d9a\u0db1 \u0d9c\u0dab\u0db1 \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Get the target tokens </p>\n": "<p>\u0d89\u0dbd\u0d9a\u0dca\u0d9a\u0d9c\u0dad\u0da7\u0ddd\u0d9a\u0db1 \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Get the tokens with the highest probabilities </p>\n": "<p>\u0d89\u0dc4\u0dc5\u0db8\u0dc3\u0db8\u0dca\u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf\u0dc0\u0db1\u0dca \u0dc3\u0dc4\u0dd2\u0dad \u0da7\u0ddd\u0d9a\u0db1 \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Input length </p>\n": "<p>\u0d86\u0daf\u0dcf\u0db1\u0daf\u0dd2\u0d9c </p>\n",
|
||||
"<p>Lengths of the input sequences </p>\n": "<p>\u0d86\u0daf\u0dcf\u0db1\u0d85\u0db1\u0dd4\u0d9a\u0dca\u0dbb\u0db8\u0dc0\u0dbd \u0daf\u0dd2\u0d9c </p>\n",
|
||||
"<p>Load the tokenizer </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca\u0db4\u0da7\u0dc0\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Log-likelihoods of the target tokens </p>\n": "<p>\u0d89\u0dbd\u0d9a\u0dca\u0d9a\u0d9c\u0dad\u0da7\u0ddd\u0d9a\u0db1 \u0dc0\u0dbd \u0dbd\u0ddc\u0d9c\u0dca \u0dc0\u0dd3\u0db8\u0dda \u0dc4\u0dd0\u0d9a\u0dd2\u0dba\u0dcf\u0dc0 </p>\n",
|
||||
"<p>Loop through each request in the chunk and collect them into PyTorch tensors with paddings </p>\n": "<p>\u0d9a\u0dd4\u0da7\u0dca\u0da7\u0dd2\u0dba\u0dda\u0d87\u0dad\u0dd2 \u0d91\u0d9a\u0dca \u0d91\u0d9a\u0dca \u0d89\u0dbd\u0dca\u0dbd\u0dd3\u0db8 \u0dc4\u0dbb\u0dc4\u0dcf \u0dbd\u0dd6\u0db4\u0dca \u0d9a\u0dbb \u0d92\u0dc0\u0dcf \u0db4\u0dd1\u0da9\u0dd2\u0db1\u0dca \u0dc3\u0db8\u0d9f \u0db4\u0dba\u0dd2\u0da7\u0ddd\u0da0\u0dca \u0da7\u0dd9\u0db1\u0dca\u0dc3\u0dbb\u0dca\u0dc0\u0dbd\u0da7 \u0d91\u0d9a\u0dad\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Loop through requests with <span translate=no>_^_0_^_</span> number of requests at a time </p>\n": "<p>\u0dc0\u0dbb\u0d9a\u0da7\u0d89\u0dbd\u0dca\u0dbd\u0dd3\u0db8\u0dca <span translate=no>_^_0_^_</span> \u0d9c\u0dab\u0db1\u0dcf\u0dc0\u0d9a\u0dca \u0dc3\u0dc4\u0dd2\u0dad \u0d89\u0dbd\u0dca\u0dbd\u0dd3\u0db8\u0dca \u0dc4\u0dbb\u0dc4\u0dcf \u0dba\u0dd0\u0dc0\u0dd3\u0db8\u0d9a\u0dca </p>\n",
|
||||
"<p>Loop through the input/output pairs of the batch </p>\n": "<p>\u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8\u0dda\u0d86\u0daf\u0dcf\u0db1/\u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1 \u0dba\u0dd4\u0d9c\u0dbd \u0dc4\u0dbb\u0dc4\u0dcf \u0dbd\u0dd6\u0db4 </p>\n",
|
||||
"<p>Maximum number of tokens to generate </p>\n": "<p>\u0d8b\u0dad\u0dca\u0db4\u0dcf\u0daf\u0db1\u0dba\u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0da7 \u0d8b\u0db4\u0dbb\u0dd2\u0db8 \u0da7\u0ddd\u0d9a\u0db1 \u0d9c\u0dab\u0db1 </p>\n",
|
||||
"<p>Maximum sequence length </p>\n": "<p>\u0d8b\u0db4\u0dbb\u0dd2\u0db8\u0d85\u0db1\u0dd4\u0d9a\u0dca\u0dbb\u0db8\u0dba \u0daf\u0dd2\u0d9c </p>\n",
|
||||
"<p>Padded length for the batch </p>\n": "<p>\u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8\u0dc3\u0db3\u0dc4\u0dcf \u0db4\u0dd1\u0da9\u0dca \u0daf\u0dd2\u0d9c </p>\n",
|
||||
"<p>Padding </p>\n": "<p>\u0db4\u0dd1\u0da9\u0dd2\u0db1\u0dca </p>\n",
|
||||
"<p>Re-order and return results </p>\n": "<p>\u0db1\u0dd0\u0dc0\u0dad\u0d87\u0dab\u0dc0\u0dd4\u0db8\u0dca \u0d9a\u0dbb \u0db1\u0dd0\u0dc0\u0dad \u0db4\u0dca\u0dbb\u0dad\u0dd2. \u0dbd </p>\n",
|
||||
"<p>Remove final token </p>\n": "<p>\u0d85\u0dc0\u0dc3\u0dcf\u0db1\u0da7\u0ddd\u0d9a\u0db1\u0dba \u0d89\u0dc0\u0dad\u0dca \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Reorder the requests in the descending order of the lengths, so that sequences with similar lengths are close </p>\n": "<p>\u0daf\u0dd2\u0d9c\u0db4\u0dc4\u0dc5 \u0d85\u0db1\u0dd4\u0db4\u0dd2\u0dc5\u0dd2\u0dc0\u0dd9\u0dbd\u0dd9\u0dc4\u0dd2 \u0d89\u0dbd\u0dca\u0dbd\u0dd3\u0db8\u0dca \u0db1\u0dd0\u0dc0\u0dad \u0dc3\u0d9a\u0dc3\u0dca \u0d9a\u0dbb\u0db1\u0dca\u0db1, \u0d91\u0dc0\u0dd2\u0da7 \u0dc3\u0db8\u0dcf\u0db1 \u0daf\u0dd2\u0d9c \u0dc3\u0dc4\u0dd2\u0dad \u0d85\u0db1\u0dd4\u0d9a\u0dca\u0dbb\u0db8\u0dba\u0db1\u0dca \u0dc3\u0db8\u0dd3\u0db4 \u0dc0\u0dda </p>\n",
|
||||
"<p>Run </p>\n": "<p>\u0daf\u0dd4\u0dc0\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Run <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">EleutherAI/lm-evaluation-harness</a> evaluator </p>\n": "<p><a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">Eleutherai/LM \u0d87\u0d9c\u0dba\u0dd3\u0db8-\u0db4\u0da7\u0dd2</a> \u0d87\u0d9c\u0dba\u0dd4\u0db8\u0dca\u0d9a\u0dbb\u0dd4 \u0db0\u0dcf\u0dc0\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Size of the vocabulary </p>\n": "<p>\u0dc0\u0da0\u0db1\u0db8\u0dcf\u0dbd\u0dcf\u0dc0\u0dda \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba </p>\n",
|
||||
"<p>The continuations for the batch </p>\n": "<p>\u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8\u0dc3\u0db3\u0dc4\u0dcf \u0d85\u0d9b\u0dab\u0dca\u0da9\u0dc0 </p>\n",
|
||||
"<p>To store the inputs for the batch </p>\n": "<p>\u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8\u0dc3\u0db3\u0dc4\u0dcf \u0dba\u0dd9\u0daf\u0dc0\u0dd4\u0db8\u0dca \u0d9c\u0db6\u0da9\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 </p>\n",
|
||||
"<p>Truncate from left if the size exceeds the <span translate=no>_^_0_^_</span> </p>\n": "<p>\u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba\u0d89\u0d9a\u0dca\u0db8\u0dc0\u0dcf \u0d9c\u0dd2\u0dba\u0dc4\u0ddc\u0dad\u0dca \u0dc0\u0db8\u0dda \u0dc3\u0dd2\u0da7 \u0d9a\u0db4\u0dcf \u0d9c\u0db1\u0dca\u0db1 <span translate=no>_^_0_^_</span> </p>\n",
|
||||
"<p>Whether there's an exact match </p>\n": "<p>\u0db1\u0dd2\u0dc1\u0dca\u0da0\u0dd2\u0dad\u0d9c\u0dd0\u0dbd\u0db4\u0dd3\u0db8\u0d9a\u0dca \u0dad\u0dd2\u0db6\u0dda\u0daf \u0dba\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>padded_length = padded_length if padded_length is not None else inplen </p>\n": "<p>padded_length= padded_length \u0db1\u0db8\u0dca padded_length \u0dc0\u0dd9\u0db1 \u0d9a\u0dd2\u0dc3\u0dd2\u0dc0\u0d9a\u0dca \u0db1\u0dd0\u0dad </p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is model </li>\n<li><span translate=no>_^_1_^_</span> is the <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a> </li>\n<li><span translate=no>_^_2_^_</span> is the size of the vocabulary (this differs from the tokenizer vocab size since neox adds some extra to make the embedding layer model parallel.) </li>\n<li><span translate=no>_^_3_^_</span> is the batch size </li>\n<li><span translate=no>_^_4_^_</span> is the device of the model</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0dba\u0db1\u0dd4 <a href=\"huggingface/tokenizers\">\u0dc4\u0d9c\u0dd2\u0d82\u0dc6\u0dda\u0dc3\u0dca \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca</a> \u0dba </li>\n<li><span translate=no>_^_2_^_</span> \u0dba\u0db1\u0dd4 \u0dc0\u0da0\u0db1 \u0db8\u0dcf\u0dbd\u0dcf\u0dc0\u0dda \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba\u0dba\u0dd2 (\u0db8\u0dd9\u0dba \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca \u0dc0\u0ddc\u0d9a\u0dcf\u0db6\u0dca \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba\u0da7 \u0dc0\u0da9\u0dcf \u0dc0\u0dd9\u0db1\u0dc3\u0dca \u0dc0\u0db1\u0dca\u0db1\u0dda \u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0d9a\u0dcf\u0dc0\u0dd0\u0daf\u0dca\u0daf\u0dd3\u0db8\u0dda \u0dc3\u0dca\u0dae\u0dbb \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb\u0dc0 \u0dc3\u0dd2\u0daf\u0dd4 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0d85\u0db8\u0dad\u0dbb \u0d85\u0db8\u0dad\u0dbb \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba\u0d9a\u0dca \u0d91\u0d9a\u0dad\u0dd4 \u0d9a\u0dbb\u0db1 \u0db6\u0dd0\u0dc0\u0dd2\u0db1\u0dd2.) </li>\n<li><span translate=no>_^_3_^_</span> \u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8 \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba \u0dc0\u0dda </li>\n</ul><li><span translate=no>_^_4_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0dda \u0d8b\u0db4\u0dcf\u0d82\u0d9c\u0dba \u0dc0\u0dda</li>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a> </li>\n<li><span translate=no>_^_1_^_</span> is the size of the vocabulary (this differs from the tokenizer vocab size since neox adds some extra to make the embedding layer model parallel.) </li>\n<li><span translate=no>_^_2_^_</span> is the batch size</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span> \u0dba\u0db1\u0dd4 <a href=\"huggingface/tokenizers\">\u0dc4\u0d9c\u0dd2\u0d82\u0dc6\u0dda\u0dc3\u0dca \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca</a> \u0dba </li>\n<li><span translate=no>_^_1_^_</span> \u0dba\u0db1\u0dd4 \u0dc0\u0da0\u0db1 \u0db8\u0dcf\u0dbd\u0dcf\u0dc0\u0dda \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba\u0dba\u0dd2 (\u0db8\u0dd9\u0dba \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca \u0dc0\u0ddc\u0d9a\u0dcf\u0db6\u0dca \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba\u0da7 \u0dc0\u0da9\u0dcf \u0dc0\u0dd9\u0db1\u0dc3\u0dca \u0dc0\u0db1\u0dca\u0db1\u0dda \u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0d9a\u0dcf\u0dc0\u0dd0\u0daf\u0dca\u0daf\u0dd3\u0db8\u0dda \u0dc3\u0dca\u0dae\u0dbb \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb\u0dc0 \u0dc3\u0dd2\u0daf\u0dd4 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0d85\u0db8\u0dad\u0dbb \u0d85\u0db8\u0dad\u0dbb \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba\u0d9a\u0dca \u0d91\u0d9a\u0dad\u0dd4 \u0d9a\u0dbb\u0db1 \u0db6\u0dd0\u0dc0\u0dd2\u0db1\u0dd2.) </li>\n<li><span translate=no>_^_2_^_</span> \u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8 \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba \u0dc0\u0dda</li></ul>\n",
|
||||
"Code to evaluate the model on NLP tasks through lm-evaluation-harness": "Lm-\u0d87\u0d9c\u0dba\u0dd3\u0db8-\u0db4\u0da7\u0dd2 \u0dc4\u0dbb\u0dc4\u0dcf NLP \u0d9a\u0dcf\u0dbb\u0dca\u0dba\u0dba\u0db1\u0dca \u0db4\u0dd2\u0dc5\u0dd2\u0db6\u0db3 \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0d87\u0d9c\u0dba\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0d9a\u0dda\u0dad\u0dba",
|
||||
"Evaluation": "\u0d87\u0d9c\u0dba\u0dd3\u0db8"
|
||||
}
|
||||
@@ -0,0 +1,54 @@
|
||||
{
|
||||
"<h1>Evaluation</h1>\n<p>This is the code to test the model on <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">EleutherAI/lm-evaluation-harness</a>.</p>\n<ul><li><a href=\"half_precision.html\">Evaluating half precision model on a single GPU</a></li></ul>\n": "<h1>\u8bc4\u4f30</h1>\n<p>\u8fd9\u662f\u5728 Ele <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">utherai/LM-Evaluation-Harnes</a> s \u4e0a\u6d4b\u8bd5\u6a21\u578b\u7684\u4ee3\u7801\u3002</p>\n<ul><li><a href=\"half_precision.html\">\u5728\u5355\u4e2a GPU \u4e0a\u8bc4\u4f30\u534a\u7cbe\u5ea6\u6a21\u578b</a></li></ul>\n",
|
||||
"<h2>Evaluation Harness Adapter</h2>\n<p>This is based on the <a href=\"https://github.com/EleutherAI/gpt-neox/blob/main/eval_tasks/eval_adapter.py\">adapter from EleutherAI/gpt-neox</a></p>\n": "<h2>\u8bc4\u4f30\u7ebf\u675f\u9002\u914d\u5668</h2>\n<p>\u8fd9\u662f\u57fa\u4e8e ele <a href=\"https://github.com/EleutherAI/gpt-neox/blob/main/eval_tasks/eval_adapter.py\">utherai/GPT-NEOX \u7684\u9002\u914d\u5668</a></p>\n",
|
||||
"<h2>Run evaluation harness with a given model</h2>\n": "<h2>\u4f7f\u7528\u7ed9\u5b9a\u6a21\u578b\u8fd0\u884c\u8bc4\u4f30\u5de5\u5177</h2>\n",
|
||||
"<h3>Get log-likelihoods of the next tokens</h3>\n<ul><li><span translate=no>_^_0_^_</span> List of requests containing the context and the expected continuation. </li>\n<li><span translate=no>_^_1_^_</span> If True, disable tqdm progress bar.</li></ul>\n": "<h3>\u83b7\u53d6\u4e0b\u4e00\u4e2a\u4ee3\u5e01\u7684\u5bf9\u6570\u53ef\u80fd\u6027</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u5305\u542b\u4e0a\u4e0b\u6587\u548c\u9884\u671f\u5ef6\u7eed\u7684\u8bf7\u6c42\u5217\u8868\u3002</li>\n<li><span translate=no>_^_1_^_</span>\u5982\u679c\u4e3a True\uff0c\u5219\u7981\u7528 tqdm \u8fdb\u5ea6\u6761\u3002</li></ul>\n",
|
||||
"<h3>Run given evaluations</h3>\n": "<h3>\u8fd0\u884c\u7ed9\u5b9a\u7684\u8bc4\u4f30</h3>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p> Batch size</p>\n": "<p>\u6279\u91cf\u5927\u5c0f</p>\n",
|
||||
"<p> Call the model</p>\n": "<p>\u7ed9\u6a21\u7279\u6253\u7535\u8bdd</p>\n",
|
||||
"<p> Decode text from token ids</p>\n": "<p>\u89e3\u7801\u6765\u81ea\u4ee4\u724c ID \u7684\u6587\u672c</p>\n",
|
||||
"<p> Encode a given text</p>\n": "<p>\u5bf9\u7ed9\u5b9a\u6587\u672c\u8fdb\u884c\u7f16\u7801</p>\n",
|
||||
"<p>Add configs </p>\n": "<p>\u6dfb\u52a0\u914d\u7f6e</p>\n",
|
||||
"<p>Add padding </p>\n": "<p>\u6dfb\u52a0\u586b\u5145</p>\n",
|
||||
"<p>Add the total log-likelihoods and whether there was a match to the results </p>\n": "<p>\u5c06\u603b\u5bf9\u6570\u4f3c\u7136\u4ee5\u53ca\u7ed3\u679c\u662f\u5426\u5b58\u5728\u5339\u914d\u9879\u76f8\u52a0</p>\n",
|
||||
"<p>All tasks if nothing is specified </p>\n": "<p>\u5982\u679c\u672a\u6307\u5b9a\u4efb\u4f55\u5185\u5bb9\uff0c\u5219\u4e3a\u6240\u6709\u4efb\u52a1</p>\n",
|
||||
"<p>Concatenate the context and continuation </p>\n": "<p>\u8fde\u63a5\u4e0a\u4e0b\u6587\u548c\u5ef6\u7eed</p>\n",
|
||||
"<p>Create a tensor </p>\n": "<p>\u521b\u5efa\u5f20\u91cf</p>\n",
|
||||
"<p>Create the adapter </p>\n": "<p>\u521b\u5efa\u9002\u914d\u5668</p>\n",
|
||||
"<p>Determine the padded length. Shorter sequences will get padded. </p>\n": "<p>\u786e\u5b9a\u586b\u5145\u7684\u957f\u5ea6\u3002\u8f83\u77ed\u7684\u5e8f\u5217\u5c06\u88ab\u586b\u5145\u3002</p>\n",
|
||||
"<p>End-of-text token </p>\n": "<p>\u6587\u672c\u7ed3\u5c3e\u4ee4\u724c</p>\n",
|
||||
"<p>For results </p>\n": "<p>\u4e3a\u4e86\u7ed3\u679c</p>\n",
|
||||
"<p>Get log softmaxes </p>\n": "<p>\u83b7\u53d6\u65e5\u5fd7 softmaxes</p>\n",
|
||||
"<p>Get logits of those </p>\n": "<p>\u83b7\u53d6\u8fd9\u4e9b\u65e5\u5fd7</p>\n",
|
||||
"<p>Get model logits </p>\n": "<p>\u83b7\u53d6\u6a21\u578b\u65e5\u5fd7</p>\n",
|
||||
"<p>Get number of predicted tokens </p>\n": "<p>\u83b7\u53d6\u9884\u6d4b\u7684\u4ee3\u5e01\u6570\u91cf</p>\n",
|
||||
"<p>Get the target tokens </p>\n": "<p>\u83b7\u53d6\u76ee\u6807\u4ee3\u5e01</p>\n",
|
||||
"<p>Get the tokens with the highest probabilities </p>\n": "<p>\u83b7\u5f97\u6982\u7387\u6700\u9ad8\u7684\u4ee3\u5e01</p>\n",
|
||||
"<p>Input length </p>\n": "<p>\u8f93\u5165\u957f\u5ea6</p>\n",
|
||||
"<p>Lengths of the input sequences </p>\n": "<p>\u8f93\u5165\u5e8f\u5217\u7684\u957f\u5ea6</p>\n",
|
||||
"<p>Load the tokenizer </p>\n": "<p>\u52a0\u8f7d\u5206\u8bcd\u5668</p>\n",
|
||||
"<p>Log-likelihoods of the target tokens </p>\n": "<p>\u76ee\u6807\u4ee3\u5e01\u7684\u5bf9\u6570\u53ef\u80fd\u6027</p>\n",
|
||||
"<p>Loop through each request in the chunk and collect them into PyTorch tensors with paddings </p>\n": "<p>\u5faa\u73af\u904d\u5386\u533a\u5757\u4e2d\u7684\u6bcf\u4e2a\u8bf7\u6c42\uff0c\u5e76\u5c06\u5b83\u4eec\u6536\u96c6\u5230\u5e26\u586b\u5145\u7684 PyTorch \u5f20\u91cf\u4e2d</p>\n",
|
||||
"<p>Loop through requests with <span translate=no>_^_0_^_</span> number of requests at a time </p>\n": "<p>\u5faa\u73af\u6d4f\u89c8\u4e00\u6b21\u5305\u542b<span translate=no>_^_0_^_</span>\u591a\u4e2a\u8bf7\u6c42\u7684\u8bf7\u6c42</p>\n",
|
||||
"<p>Loop through the input/output pairs of the batch </p>\n": "<p>\u5faa\u73af\u6d4f\u89c8\u6279\u6b21\u7684\u8f93\u5165/\u8f93\u51fa\u5bf9</p>\n",
|
||||
"<p>Maximum number of tokens to generate </p>\n": "<p>\u8981\u751f\u6210\u7684\u4ee4\u724c\u7684\u6700\u5927\u6570\u91cf</p>\n",
|
||||
"<p>Maximum sequence length </p>\n": "<p>\u6700\u5927\u5e8f\u5217\u957f\u5ea6</p>\n",
|
||||
"<p>Padded length for the batch </p>\n": "<p>\u6279\u6b21\u7684\u586b\u5145\u957f\u5ea6</p>\n",
|
||||
"<p>Padding </p>\n": "<p>\u586b\u5145</p>\n",
|
||||
"<p>Re-order and return results </p>\n": "<p>\u91cd\u65b0\u6392\u5e8f\u5e76\u8fd4\u56de\u7ed3\u679c</p>\n",
|
||||
"<p>Remove final token </p>\n": "<p>\u79fb\u9664\u6700\u7ec8\u4ee4\u724c</p>\n",
|
||||
"<p>Reorder the requests in the descending order of the lengths, so that sequences with similar lengths are close </p>\n": "<p>\u6309\u957f\u5ea6\u7684\u964d\u5e8f\u5bf9\u8bf7\u6c42\u8fdb\u884c\u91cd\u65b0\u6392\u5e8f\uff0c\u4ee5\u4f7f\u957f\u5ea6\u76f8\u4f3c\u7684\u5e8f\u5217\u63a5\u8fd1</p>\n",
|
||||
"<p>Run </p>\n": "<p>\u8dd1</p>\n",
|
||||
"<p>Run <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">EleutherAI/lm-evaluation-harness</a> evaluator </p>\n": "<p>\u8fd0\u884c <a href=\"https://github.com/EleutherAI/lm-evaluation-harness\">eleutherai/LM-Evaluation-Harnes</a> s \u8bc4\u4f30\u5668</p>\n",
|
||||
"<p>Size of the vocabulary </p>\n": "<p>\u8bcd\u6c47\u91cf\u7684\u5927\u5c0f</p>\n",
|
||||
"<p>The continuations for the batch </p>\n": "<p>\u8be5\u6279\u6b21\u7684\u5ef6\u7eed</p>\n",
|
||||
"<p>To store the inputs for the batch </p>\n": "<p>\u5b58\u50a8\u6279\u6b21\u7684\u8f93\u5165</p>\n",
|
||||
"<p>Truncate from left if the size exceeds the <span translate=no>_^_0_^_</span> </p>\n": "<p>\u5982\u679c\u5927\u5c0f\u8d85\u8fc7<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Whether there's an exact match </p>\n": "<p>\u662f\u5426\u5b58\u5728\u5b8c\u5168\u5339\u914d</p>\n",
|
||||
"<p>padded_length = padded_length if padded_length is not None else inplen </p>\n": "<p>\u5982\u679c padded_length \u4e0d\u662f padded_length \u5219\u4e3a padded_length \u5176\u4ed6\u6ca1\u6709 inplen</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is model </li>\n<li><span translate=no>_^_1_^_</span> is the <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a> </li>\n<li><span translate=no>_^_2_^_</span> is the size of the vocabulary (this differs from the tokenizer vocab size since neox adds some extra to make the embedding layer model parallel.) </li>\n<li><span translate=no>_^_3_^_</span> is the batch size </li>\n<li><span translate=no>_^_4_^_</span> is the device of the model</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u6a21\u7279</li>\n<li><span translate=no>_^_1_^_</span>\u662f <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a></li>\n<li><span translate=no>_^_2_^_</span>\u662f\u8bcd\u6c47\u91cf\u7684\u5927\u5c0f\uff08\u8fd9\u4e0e\u5206\u8bcd\u5668\u8bcd\u6c47\u5927\u5c0f\u4e0d\u540c\uff0c\u56e0\u4e3aneox\u6dfb\u52a0\u4e86\u4e00\u4e9b\u989d\u5916\u7684\u5185\u5bb9\u6765\u4f7f\u5d4c\u5165\u5c42\u6a21\u578b\u5e76\u884c\u3002\uff09</li>\n<li><span translate=no>_^_3_^_</span>\u662f\u6279\u6b21\u5927\u5c0f</li>\n<li><span translate=no>_^_4_^_</span>\u662f\u8be5\u578b\u53f7\u7684\u8bbe\u5907</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a> </li>\n<li><span translate=no>_^_1_^_</span> is the size of the vocabulary (this differs from the tokenizer vocab size since neox adds some extra to make the embedding layer model parallel.) </li>\n<li><span translate=no>_^_2_^_</span> is the batch size</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f <a href=\"huggingface/tokenizers\">Huggingface Tokenizer</a></li>\n<li><span translate=no>_^_1_^_</span>\u662f\u8bcd\u6c47\u91cf\u7684\u5927\u5c0f\uff08\u8fd9\u4e0e\u5206\u8bcd\u5668\u8bcd\u6c47\u5927\u5c0f\u4e0d\u540c\uff0c\u56e0\u4e3aneox\u6dfb\u52a0\u4e86\u4e00\u4e9b\u989d\u5916\u7684\u5185\u5bb9\u6765\u4f7f\u5d4c\u5165\u5c42\u6a21\u578b\u5e76\u884c\u3002\uff09</li>\n<li><span translate=no>_^_2_^_</span>\u662f\u6279\u6b21\u5927\u5c0f</li></ul>\n",
|
||||
"Code to evaluate the model on NLP tasks through lm-evaluation-harness": "\u901a\u8fc7 lm-evaluation-harness \u8bc4\u4f30\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\u6a21\u578b\u7684\u4ee3\u7801",
|
||||
"Evaluation": "\u8bc4\u4f30"
|
||||
}
|
||||
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"<h1>Evaluate GPT-NeoX using LLM.int8() quantization on test suite</h1>\n<p>This code evaluate <a href=\"../index.html\">GPT-NeoX</a> using, on a suite of tasks.</p>\n": "<h1>\u30c6\u30b9\u30c8\u30b9\u30a4\u30fc\u30c8\u3067 llm.int8 () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066 GPT-Neox \u3092\u8a55\u4fa1\u3059\u308b</h1>\n<p>\u3053\u306e\u30b3\u30fc\u30c9\u3067\u306f\u3001<a href=\"../index.html\">\u4e00\u9023\u306e\u30bf\u30b9\u30af\u3067 GPT-Neox</a> \u3092\u4f7f\u7528\u3057\u3066\u8a55\u4fa1\u3057\u307e\u3059\u3002</p>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Argument parser </p>\n": "<p>\u5f15\u6570\u30d1\u30fc\u30b5\u30fc</p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p><span translate=no>_^_0_^_</span>\u30e2\u30c7\u30eb\u4f5c\u6210</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u7aef\u672b</p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u30ec\u30a4\u30e4\u30fc\u3092\u30ed\u30fc\u30c9</p>\n",
|
||||
"<p>Run <a href=\"index.html\">evaluation harness</a> </p>\n": "<p><a href=\"index.html\">\u8a55\u4fa1\u7528\u30cf\u30fc\u30cd\u30b9\u3092\u5b9f\u884c</a></p>\n",
|
||||
"Evaluate GPT-NeoX using LLM.int8() quantization on test suite": "\u30c6\u30b9\u30c8\u30b9\u30a4\u30fc\u30c8\u3067 llm.int8 () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066 GPT-Neox \u3092\u8a55\u4fa1\u3059\u308b"
|
||||
}
|
||||
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"<h1>Evaluate GPT-NeoX using LLM.int8() quantization on test suite</h1>\n<p>This code evaluate <a href=\"../index.html\">GPT-NeoX</a> using, on a suite of tasks.</p>\n": "<h1>LLM.INT8\u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dad\u0d9a\u0dca\u0dc3\u0dda\u0dbb\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 () \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dab \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba \u0db8\u0dad \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba</h1>\n<p>\u0db8\u0dd9\u0db8\u0d9a\u0dda\u0dad\u0dba \u0d9a\u0dcf\u0dbb\u0dca\u0dba\u0dba\u0db1\u0dca \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba\u0d9a\u0dd2\u0db1\u0dca \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca <a href=\"../index.html\">\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca</a> \u0d87\u0d9c\u0dba\u0dd3\u0db8\u0da7 \u0dbd\u0d9a\u0dca \u0d9a\u0dbb\u0dba\u0dd2. </p>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Argument parser </p>\n": "<p>\u0dad\u0dbb\u0dca\u0d9a \u0dc0\u0dd2\u0dad\u0dbb\u0dca\u0d9a</p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p><span translate=no>_^_0_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Device </p>\n": "<p>\u0d8b\u0db4\u0dcf\u0d82\u0d9c\u0dba </p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u0dc3\u0dca\u0dae\u0dbb\u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Run <a href=\"index.html\">evaluation harness</a> </p>\n": "<p><a href=\"index.html\">\u0d87\u0d9c\u0dba\u0dd3\u0db8\u0dca \u0db4\u0da7\u0dd2</a> \u0db0\u0dcf\u0dc0\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"Evaluate GPT-NeoX using LLM.int8() quantization on test suite": "LLM.INT8 \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dad\u0d9a\u0dca\u0dc3\u0dda\u0dbb\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 () \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dab \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba \u0db8\u0dad \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba"
|
||||
}
|
||||
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"<h1>Evaluate GPT-NeoX using LLM.int8() quantization on test suite</h1>\n<p>This code evaluate <a href=\"../index.html\">GPT-NeoX</a> using, on a suite of tasks.</p>\n": "<h1>\u5728\u6d4b\u8bd5\u5957\u4ef6\u4e0a\u4f7f\u7528 llm.int8 () \u91cf\u5316\u6765\u8bc4\u4f30 GPT-NEOX</h1>\n<p>\u6b64\u4ee3\u7801\u4f7f\u7528\u5728\u4e00\u5957\u4efb\u52a1\u4e0a\u8bc4\u4f30 <a href=\"../index.html\">GPT-NEOX</a>\u3002</p>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Argument parser </p>\n": "<p>\u53c2\u6570\u89e3\u6790\u5668</p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p>\u521b\u5efa<span translate=no>_^_0_^_</span>\u6a21\u578b</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u8bbe\u5907</p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u52a0\u8f7d\u56fe\u5c42</p>\n",
|
||||
"<p>Run <a href=\"index.html\">evaluation harness</a> </p>\n": "<p>\u8fd0\u884c<a href=\"index.html\">\u8bc4\u4f30\u5de5\u5177</a></p>\n",
|
||||
"Evaluate GPT-NeoX using LLM.int8() quantization on test suite": "\u5728\u6d4b\u8bd5\u5957\u4ef6\u4e0a\u4f7f\u7528 llm.int8 () \u91cf\u5316\u6765\u8bc4\u4f30 GPT-NEOX"
|
||||
}
|
||||
@@ -0,0 +1,11 @@
|
||||
{
|
||||
"<h1>Evaluate GPT-NeoX using LLM.int8() quantization on test suite</h1>\n<p>This code evaluate <a href=\"../index.html\">GPT-NeoX</a> using <a href=\"../utils/llm_int8.html\">LLM.int8() quantization</a>, on a suite of tasks.</p>\n": "<h1>\u30c6\u30b9\u30c8\u30b9\u30a4\u30fc\u30c8\u3067 llm.int8 () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066 GPT-Neox \u3092\u8a55\u4fa1\u3059\u308b</h1>\n<p>\u3053\u306e\u30b3\u30fc\u30c9\u306f\u3001\u4e00\u9023\u306e\u30bf\u30b9\u30af\u3067 <a href=\"../utils/llm_int8.html\">llm.int8 () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066</a> <a href=\"../index.html\">GPT-Neox</a> \u3092\u8a55\u4fa1\u3057\u307e\u3059\u3002</p>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p><span translate=no>_^_0_^_</span>\u30e2\u30c7\u30eb\u4f5c\u6210</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u7aef\u672b</p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u30ec\u30a4\u30e4\u30fc\u3092\u30ed\u30fc\u30c9</p>\n",
|
||||
"<p>Load layers in float16 into CPU. We convert the layers to int8 later, because doing that on the fly after loading layers to GPU causes CUDA memory fragmentation (about 3GB memory can get lost due to fragmentation). </p>\n": "<p>float16 \u306e\u30ec\u30a4\u30e4\u30fc\u3092 CPU \u306b\u30ed\u30fc\u30c9\u3057\u307e\u3059\u3002\u30ec\u30a4\u30e4\u30fc\u3092GPU\u306b\u30ed\u30fc\u30c9\u3057\u305f\u5f8c\u306b\u305d\u306e\u5834\u3067\u3053\u308c\u3092\u884c\u3046\u3068\u3001CUDA\u30e1\u30e2\u30ea\u306e\u65ad\u7247\u5316\u304c\u767a\u751f\u3059\u308b\u305f\u3081\u3001\u5f8c\u3067\u30ec\u30a4\u30e4\u30fc\u3092int8\u306b\u5909\u63db\u3057\u307e\u3059\uff08\u30d5\u30e9\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u306b\u3088\u308a\u7d043GB\u306e\u30e1\u30e2\u30ea\u304c\u5931\u308f\u308c\u308b\u53ef\u80fd\u6027\u304c\u3042\u308a\u307e\u3059</p>\uff09\u3002\n",
|
||||
"<p>Run <a href=\"index.html\">evaluation harness</a> </p>\n": "<p><a href=\"index.html\">\u8a55\u4fa1\u7528\u30cf\u30fc\u30cd\u30b9\u3092\u5b9f\u884c</a></p>\n",
|
||||
"<p>This reduces CUDA memory fragmentation </p>\n": "<p>\u3053\u308c\u306b\u3088\u308a\u3001CUDA \u30e1\u30e2\u30ea\u306e\u65ad\u7247\u5316\u304c\u6e1b\u5c11\u3057\u307e\u3059\u3002</p>\n",
|
||||
"Evaluate GPT-NeoX using LLM.int8() quantization on test suite": "\u30c6\u30b9\u30c8\u30b9\u30a4\u30fc\u30c8\u3067 llm.int8 () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066 GPT-Neox \u3092\u8a55\u4fa1\u3059\u308b"
|
||||
}
|
||||
@@ -0,0 +1,11 @@
|
||||
{
|
||||
"<h1>Evaluate GPT-NeoX using LLM.int8() quantization on test suite</h1>\n<p>This code evaluate <a href=\"../index.html\">GPT-NeoX</a> using <a href=\"../utils/llm_int8.html\">LLM.int8() quantization</a>, on a suite of tasks.</p>\n": "<h1>LLM.INT8\u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dad\u0d9a\u0dca\u0dc3\u0dda\u0dbb\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 () \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dab \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba \u0db8\u0dad \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba</h1>\n<p>\u0db8\u0dd9\u0db8\u0d9a\u0dda\u0dad\u0dba <a href=\"../utils/llm_int8.html\">LLM.INT8 () \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba</a> \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca <a href=\"../index.html\">\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca</a> \u0d87\u0d9c\u0dba\u0dd3\u0db8\u0da7 \u0dbd\u0d9a\u0dca \u0d9a\u0dbb\u0dba\u0dd2, \u0d9a\u0dcf\u0dbb\u0dca\u0dba\u0dba\u0db1\u0dca \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba\u0d9a\u0dca \u0db8\u0dad. </p>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p><span translate=no>_^_0_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Device </p>\n": "<p>\u0d8b\u0db4\u0dcf\u0d82\u0d9c\u0dba </p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u0dc3\u0dca\u0dae\u0dbb\u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Load layers in float16 into CPU. We convert the layers to int8 later, because doing that on the fly after loading layers to GPU causes CUDA memory fragmentation (about 3GB memory can get lost due to fragmentation). </p>\n": "<p>\u0db4\u0dcf\u0dc0\u0dd9\u0db116 \u0dc4\u0dd2 \u0dc3\u0dca\u0dae\u0dbb CPU \u0dad\u0dd4\u0dc5\u0da7 \u0db4\u0da7\u0dc0\u0db1\u0dca\u0db1. \u0d85\u0db4\u0dd2 \u0dc3\u0dca\u0dae\u0dbb \u0db4\u0dc3\u0dd4\u0dc0 int8 \u0db6\u0dc0\u0da7 \u0db4\u0dbb\u0dd2\u0dc0\u0dbb\u0dca\u0dad\u0db1\u0dba \u0d9a\u0dbb\u0db8\u0dd4, \u0db8\u0db1\u0dca\u0daf \u0dc3\u0dca\u0dae\u0dbb GPU \u0dc0\u0dd9\u0dad \u0db4\u0dd0\u0da7\u0dc0\u0dd3\u0db8\u0dd9\u0db1\u0dca \u0db4\u0dc3\u0dd4 \u0db4\u0dd2\u0dba\u0dcf\u0dc3\u0dbb \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 CUDA \u0db8\u0dad\u0d9a \u0d9b\u0dab\u0dca\u0da9\u0db1\u0dba \u0dc0\u0dd3\u0db8\u0da7 \u0dc4\u0dda\u0dad\u0dd4 \u0dc0\u0dda (3GB \u0db4\u0db8\u0dab \u0db8\u0dad\u0d9a\u0dba \u0d9a\u0dd0\u0db6\u0dbd\u0dd2 \u0dc0\u0dd3\u0db8 \u0db1\u0dd2\u0dc3\u0dcf \u0d85\u0dc4\u0dd2\u0db8\u0dd2 \u0dc0\u0dd2\u0dba \u0dc4\u0dd0\u0d9a). </p>\n",
|
||||
"<p>Run <a href=\"index.html\">evaluation harness</a> </p>\n": "<p><a href=\"index.html\">\u0d87\u0d9c\u0dba\u0dd3\u0db8\u0dca \u0db4\u0da7\u0dd2</a> \u0db0\u0dcf\u0dc0\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>This reduces CUDA memory fragmentation </p>\n": "<p>\u0db8\u0dd9\u0dbaCUDA \u0db8\u0dad\u0d9a \u0d9b\u0dab\u0dca\u0da9\u0db1\u0dba \u0d85\u0da9\u0dd4 \u0d9a\u0dbb\u0dba\u0dd2 </p>\n",
|
||||
"Evaluate GPT-NeoX using LLM.int8() quantization on test suite": "LLM.INT8 \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dad\u0d9a\u0dca\u0dc3\u0dda\u0dbb\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 () \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dab \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba \u0db8\u0dad \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba"
|
||||
}
|
||||
@@ -0,0 +1,11 @@
|
||||
{
|
||||
"<h1>Evaluate GPT-NeoX using LLM.int8() quantization on test suite</h1>\n<p>This code evaluate <a href=\"../index.html\">GPT-NeoX</a> using <a href=\"../utils/llm_int8.html\">LLM.int8() quantization</a>, on a suite of tasks.</p>\n": "<h1>\u5728\u6d4b\u8bd5\u5957\u4ef6\u4e0a\u4f7f\u7528 llm.int8 () \u91cf\u5316\u6765\u8bc4\u4f30 GPT-NEOX</h1>\n<p>\u8fd9\u6bb5\u4ee3\u7801\u5728\u4e00\u5957\u4efb\u52a1\u4e2d\u4f7f\u7528 <a href=\"../utils/llm_int8.html\">llm.int8 () \u91cf\u5316</a>\u6765\u8bc4\u4f30 <a href=\"../index.html\">GPT-NEOX</a>\u3002</p>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p>\u521b\u5efa<span translate=no>_^_0_^_</span>\u6a21\u578b</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u8bbe\u5907</p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u52a0\u8f7d\u56fe\u5c42</p>\n",
|
||||
"<p>Load layers in float16 into CPU. We convert the layers to int8 later, because doing that on the fly after loading layers to GPU causes CUDA memory fragmentation (about 3GB memory can get lost due to fragmentation). </p>\n": "\u5c06 <p>float16 \u4e2d\u7684\u5c42\u52a0\u8f7d\u5230 CPU \u4e2d\u3002\u6211\u4eec\u7a0d\u540e\u5c06\u56fe\u5c42\u8f6c\u6362\u4e3aint8\uff0c\u56e0\u4e3a\u5728\u5c06\u56fe\u5c42\u52a0\u8f7d\u5230GPU\u540e\u5373\u65f6\u6267\u884c\u6b64\u64cd\u4f5c\u4f1a\u5bfc\u81f4CUDA\u5185\u5b58\u788e\u7247\uff08\u5927\u7ea63GB\u7684\u5185\u5b58\u53ef\u80fd\u4f1a\u7531\u4e8e\u788e\u7247\u800c\u4e22\u5931\uff09\u3002</p>\n",
|
||||
"<p>Run <a href=\"index.html\">evaluation harness</a> </p>\n": "<p>\u8fd0\u884c<a href=\"index.html\">\u8bc4\u4f30\u5de5\u5177</a></p>\n",
|
||||
"<p>This reduces CUDA memory fragmentation </p>\n": "<p>\u8fd9\u51cf\u5c11\u4e86 CUDA \u5185\u5b58\u788e\u7247</p>\n",
|
||||
"Evaluate GPT-NeoX using LLM.int8() quantization on test suite": "\u5728\u6d4b\u8bd5\u5957\u4ef6\u4e0a\u4f7f\u7528 llm.int8 () \u91cf\u5316\u6765\u8bc4\u4f30 GPT-NEOX"
|
||||
}
|
||||
@@ -0,0 +1,113 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Model</h1>\n<p>Here is the code for layers of GPT-NeoX model and the code to load 20B checkpoint.</p>\n<p>The method <span translate=no>_^_0_^_</span> in the layers load the checkpoints of that layer. The checkpoint loading helpers are on <a href=\"checkpoint.html\"><span translate=no>_^_1_^_</span></a></p>\n": "<h1>GPT \u30cd\u30aa\u30c3\u30af\u30b9\u30e2\u30c7\u30eb</h1>\n<p>\u3053\u308c\u306f\u3001GPT-Neox\u30e2\u30c7\u30eb\u306e\u30ec\u30a4\u30e4\u30fc\u7528\u306e\u30b3\u30fc\u30c9\u306820B\u306e\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3092\u30ed\u30fc\u30c9\u3059\u308b\u30b3\u30fc\u30c9\u3067\u3059\u3002</p>\n<p><span translate=no>_^_0_^_</span>\u30ec\u30a4\u30e4\u30fc\u5185\u306e\u30e1\u30bd\u30c3\u30c9\u306f\u3001\u305d\u306e\u30ec\u30a4\u30e4\u30fc\u306e\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3092\u30ed\u30fc\u30c9\u3057\u307e\u3059\u3002\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u30ed\u30fc\u30c9\u30d8\u30eb\u30d1\u30fc\u304c\u30aa\u30f3\u306b\u306a\u3063\u3066\u3044\u307e\u3059 <a href=\"checkpoint.html\"><span translate=no>_^_1_^_</span></a></p>\n",
|
||||
"<h2>Attention layer</h2>\n": "<h2>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30ec\u30a4\u30e4\u30fc</h2>\n",
|
||||
"<h2>Embedding layer</h2>\n<p>This is a standard embeddings layer with code to load the checkpoint.</p>\n": "<h2>\u57cb\u3081\u8fbc\u307f\u30ec\u30a4\u30e4\u30fc</h2>\n<p>\u3053\u308c\u306f\u3001\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3092\u30ed\u30fc\u30c9\u3059\u308b\u30b3\u30fc\u30c9\u3092\u542b\u3080\u6a19\u6e96\u306e\u57cb\u3081\u8fbc\u307f\u30ec\u30a4\u30e4\u30fc\u3067\u3059\u3002</p>\n",
|
||||
"<h2>Feedforward Network</h2>\n": "<h2>\u30d5\u30a3\u30fc\u30c9\u30d5\u30a9\u30ef\u30fc\u30c9\u30cd\u30c3\u30c8\u30ef\u30fc\u30af</h2>\n",
|
||||
"<h2>Final normalization layer</h2>\n": "<h2>\u6700\u7d42\u6b63\u898f\u5316\u30ec\u30a4\u30e4\u30fc</h2>\n",
|
||||
"<h2>Rotary Positional Embeddings</h2>\n<p>GPT-NeoX uses <a href=\"https://arxiv.org/abs/2104.09864\">rotary positional embeddings (RoPE)</a>.</p>\n<p>WE have annotated implementation of RoPE <a href=\"https://nn.labml.ai/transformers/rope/index.html\">here</a> with more notes the theory.</p>\n": "<h2>\u30ed\u30fc\u30bf\u30ea\u30fc\u30dd\u30b8\u30b7\u30e7\u30ca\u30eb\u30a8\u30f3\u30d9\u30c7\u30a3\u30f3\u30b0</h2>\n<p><a href=\"https://arxiv.org/abs/2104.09864\">GPT-Neox\u306f\u56de\u8ee2\u5f0f\u30dd\u30b8\u30b7\u30e7\u30ca\u30eb\u30a8\u30f3\u30d9\u30c7\u30a3\u30f3\u30b0</a>\uff08RoPE\uff09\u3092\u4f7f\u7528\u3057\u3066\u3044\u307e\u3059\u3002</p>\n<p><a href=\"https://nn.labml.ai/transformers/rope/index.html\">\u3053\u3053\u3067\u306f</a>\u3001RoPE \u306e\u5b9f\u88c5\u306b\u6ce8\u91c8\u3092\u4ed8\u3051\u3066\u3001\u7406\u8ad6\u306b\u95a2\u3059\u308b\u6ce8\u91c8\u3092\u4ed8\u3051\u307e\u3057\u305f\u3002</p>\n",
|
||||
"<h2>Transformer Layer</h2>\n": "<h2>\u5909\u5727\u5668\u5c64</h2>\n",
|
||||
"<h3>Generator to create layers</h3>\n<p>The layers are generated in the same order as checkpoints.</p>\n<p>It gives <span translate=no>_^_0_^_</span> when a layer is not available; we use the layer indices as NeoX and there are two transformation layers we don't need in our implementation.</p>\n<ul><li><span translate=no>_^_1_^_</span> is the number of tokens in the vocabulary </li>\n<li><span translate=no>_^_2_^_</span> is the number of features in the embeddings </li>\n<li><span translate=no>_^_3_^_</span> is the number of transformer layers </li>\n<li><span translate=no>_^_4_^_</span> is the number of attention heads </li>\n<li><span translate=no>_^_5_^_</span> are the set of layers to be used. All layers will be used if None. This is used to test smaller versions of the model with fewer layers </li>\n<li><span translate=no>_^_6_^_</span> specifies whether to clone the transformer layers (a bit faster) </li>\n<li><span translate=no>_^_7_^_</span> is the data type of the model </li>\n<li><span translate=no>_^_8_^_</span> is the device of the model </li>\n<li><span translate=no>_^_9_^_</span> specifies whether to use int8 quantization </li>\n<li><span translate=no>_^_10_^_</span> is the threshold <span translate=no>_^_11_^_</span> used to separate outlier features </li>\n<li><span translate=no>_^_12_^_</span> specifies whether to use <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n": "<h3>\u30ec\u30a4\u30e4\u30fc\u3092\u4f5c\u6210\u3059\u308b\u305f\u3081\u306e\u30b8\u30a7\u30cd\u30ec\u30fc\u30bf\u30fc</h3>\n<p>\u30ec\u30a4\u30e4\u30fc\u306f\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3068\u540c\u3058\u9806\u5e8f\u3067\u751f\u6210\u3055\u308c\u307e\u3059\u3002</p>\n<p><span translate=no>_^_0_^_</span>\u30ec\u30a4\u30e4\u30fc\u304c\u4f7f\u7528\u3067\u304d\u306a\u3044\u5834\u5408\u306b\u8fd4\u3055\u308c\u307e\u3059\u3002\u30ec\u30a4\u30e4\u30fc\u30a4\u30f3\u30c7\u30c3\u30af\u30b9\u3092NeoX\u3068\u3057\u3066\u4f7f\u7528\u3057\u3001\u5b9f\u88c5\u306b\u306f\u5fc5\u8981\u306e\u306a\u3044\u5909\u63db\u30ec\u30a4\u30e4\u30fc\u304c2\u3064\u3042\u308a\u307e\u3059\u3002</p>\n<ul><li><span translate=no>_^_1_^_</span>\u30dc\u30ad\u30e3\u30d6\u30e9\u30ea\u5185\u306e\u30c8\u30fc\u30af\u30f3\u306e\u6570\u3067\u3059</li>\n<li><span translate=no>_^_2_^_</span>\u306f\u57cb\u3081\u8fbc\u307f\u5185\u306e\u30d5\u30a3\u30fc\u30c1\u30e3\u306e\u6570\u3067\u3059</li>\n<li><span translate=no>_^_3_^_</span>\u5909\u5727\u5668\u5c64\u306e\u6570\u3067\u3059</li>\n<li><span translate=no>_^_4_^_</span>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30fb\u30d8\u30c3\u30c9\u306e\u6570\u3067\u3059</li>\n<li><span translate=no>_^_5_^_</span>\u4f7f\u7528\u3059\u308b\u30ec\u30a4\u30e4\u30fc\u306e\u30bb\u30c3\u30c8\u3067\u3059\u3002None \u306e\u5834\u5408\u306f\u3059\u3079\u3066\u306e\u30ec\u30a4\u30e4\u30fc\u304c\u4f7f\u7528\u3055\u308c\u307e\u3059\u3002\u3053\u308c\u306f\u3001\u30ec\u30a4\u30e4\u30fc\u6570\u306e\u5c11\u306a\u3044\u30e2\u30c7\u30eb\u306e\u5c0f\u3055\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u30c6\u30b9\u30c8\u3059\u308b\u5834\u5408\u306b\u4f7f\u7528\u3057\u307e\u3059</li>\u3002\n<li><span translate=no>_^_6_^_</span>\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30ec\u30a4\u30e4\u30fc\u306e\u30af\u30ed\u30fc\u30f3\u3092\u4f5c\u6210\u3059\u308b\u304b\u3069\u3046\u304b\u3092\u6307\u5b9a\u3057\u307e\u3059 (\u5c11\u3057\u901f\u304f\u306a\u308a\u307e\u3059)</li>\n<li><span translate=no>_^_7_^_</span>\u30e2\u30c7\u30eb\u306e\u30c7\u30fc\u30bf\u578b\u3067\u3059</li>\n<li><span translate=no>_^_8_^_</span>\u30e2\u30c7\u30eb\u306e\u30c7\u30d0\u30a4\u30b9\u3067\u3059</li>\n<li><span translate=no>_^_9_^_</span>int8 \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3059\u308b\u304b\u3069\u3046\u304b\u3092\u6307\u5b9a\u3057\u307e\u3059</li>\n<li><span translate=no>_^_10_^_</span><span translate=no>_^_11_^_</span>\u5916\u308c\u5024\u306e\u7279\u5fb4\u3092\u5206\u96e2\u3059\u308b\u305f\u3081\u306e\u95be\u5024\u3067\u3059</li>\n<li><span translate=no>_^_12_^_</span><a href=\"https://github.com/HazyResearch/flash-attention\">\u30d5\u30e9\u30c3\u30b7\u30e5\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3059\u308b\u304b\u3069\u3046\u304b\u3092\u6307\u5b9a\u3057\u307e\u3059</a></li></ul>\n",
|
||||
"<h3>Generator to get layers</h3>\n": "<h3>\u30ec\u30a4\u30e4\u30fc\u3092\u53d6\u5f97\u3059\u308b\u305f\u3081\u306e\u30b8\u30a7\u30cd\u30ec\u30fc\u30bf\u30fc</h3>\n",
|
||||
"<h3>Generator to load layers</h3>\n": "<h3>\u30ec\u30a4\u30e4\u30fc\u3092\u30ed\u30fc\u30c9\u3059\u308b\u30b8\u30a7\u30cd\u30ec\u30fc\u30bf\u30fc</h3>\n",
|
||||
"<h3>Returns the total number of layers</h3>\n": "<h3>\u30ec\u30a4\u30e4\u30fc\u306e\u7dcf\u6570\u3092\u8fd4\u3057\u307e\u3059</h3>\n",
|
||||
"<h3>Rotate the features</h3>\n<p><span translate=no>_^_0_^_</span></p>\n": "<h3>\u30d5\u30a3\u30fc\u30c1\u30e3\u3092\u30ed\u30fc\u30c6\u30fc\u30b7\u30e7\u30f3\u3057\u3066\u304f\u3060\u3055\u3044</h3>\n<p><span translate=no>_^_0_^_</span></p>\n",
|
||||
"<h4>Calculate the causal mask</h4>\n<ul><li><span translate=no>_^_0_^_</span> has shape <a href=\"batch_size, query_seq_len, key_seq_len, n_heads\">batch_size, query_seq_len, key_seq_len, n_heads</a></li></ul>\n": "<h4>\u56e0\u679c\u30de\u30b9\u30af\u306e\u8a08\u7b97</h4>\n<ul><li><span translate=no>_^_0_^_</span><a href=\"batch_size, query_seq_len, key_seq_len, n_heads\">\u30d0\u30c3\u30c1\u30b5\u30a4\u30ba\u3001\u30af\u30a8\u30ea\u30b7\u30fc\u30b1\u30f3\u30b9\u30ec\u30f3\u3001\u30ad\u30fc\u30b7\u30fc\u30b1\u30f3\u30b9\u30ec\u30f3\u3001</a> n\u30d8\u30c3\u30ba\u306e\u30b7\u30a7\u30a4\u30d7\u304c\u3042\u308a\u307e\u3059</li></ul>\n",
|
||||
"<h4>Creates and caches a layer</h4>\n<p>Copying cached layers is faster than initializing new layers because it takes time to initialize parameters.</p>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the layer </li>\n<li><span translate=no>_^_1_^_</span> is the function to create the layer </li>\n<p><em>Returns</em> the created layer or a copy of the cached layer</p></ul>\n": "<h4>\u30ec\u30a4\u30e4\u30fc\u3092\u4f5c\u6210\u3057\u3066\u30ad\u30e3\u30c3\u30b7\u30e5\u3057\u307e\u3059</h4>\n<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u3055\u308c\u305f\u30ec\u30a4\u30e4\u30fc\u306e\u30b3\u30d4\u30fc\u306f\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306e\u521d\u671f\u5316\u306b\u6642\u9593\u304c\u304b\u304b\u308b\u305f\u3081\u3001\u65b0\u3057\u3044\u30ec\u30a4\u30e4\u30fc\u3092\u521d\u671f\u5316\u3059\u308b\u3088\u308a\u3082\u9ad8\u901f\u3067\u3059\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u30ec\u30a4\u30e4\u30fc\u306e\u540d\u524d\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u30ec\u30a4\u30e4\u30fc\u3092\u4f5c\u6210\u3059\u308b\u95a2\u6570\u3067\u3059</li>\n<p><em>\u4f5c\u6210\u3055\u308c\u305f\u30ec\u30a4\u30e4\u30fc\u307e\u305f\u306f\u30ad\u30e3\u30c3\u30b7\u30e5\u3055\u308c\u305f\u30ec\u30a4\u30e4\u30fc\u306e\u30b3\u30d4\u30fc\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h4>Prepares the layer for usage</h4>\n<p>We move the layer to the device and convert it to the correct data type</p>\n<ul><li><span translate=no>_^_0_^_</span> is the layer to prepare </li>\n<p><em>Returns</em> the prepared layer</p></ul>\n": "<h4>\u30ec\u30a4\u30e4\u30fc\u3092\u4f7f\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u6e96\u5099\u3057\u307e\u3059</h4>\n<p>\u30ec\u30a4\u30e4\u30fc\u3092\u30c7\u30d0\u30a4\u30b9\u306b\u79fb\u52d5\u3057\u3001\u6b63\u3057\u3044\u30c7\u30fc\u30bf\u578b\u306b\u5909\u63db\u3057\u307e\u3059\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u6e96\u5099\u3059\u308b\u30ec\u30a4\u30e4\u30fc\u3067\u3059</li>\n<p><em>\u6e96\u5099\u3057\u305f\u30ec\u30a4\u30e4\u30fc\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p> <a id=\"post_load_prepare\"></a></p>\n<h3>Layer transformations after loading the checkpoint</h3>\n<p>This function implements layer transformations after loading the checkpoint.</p>\n<p>Currently, it only applies the int8 quantization.</p>\n<ul><li><span translate=no>_^_0_^_</span> is the layer to prepare </li>\n<li><span translate=no>_^_1_^_</span> specifies whether to use int8 quantization </li>\n<li><span translate=no>_^_2_^_</span> is the device of the model </li>\n<li><span translate=no>_^_3_^_</span> is the threshold <span translate=no>_^_4_^_</span> used to separate outlier features </li>\n<p><em>Returns</em> the prepared layer</p></ul>\n": "<p><a id=\"post_load_prepare\"></a></p>\n<h3>\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3092\u30ed\u30fc\u30c9\u3057\u305f\u5f8c\u306e\u30ec\u30a4\u30e4\u30fc\u5909\u63db</h3>\n<p>\u3053\u306e\u95a2\u6570\u306f\u3001\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3092\u8aad\u307f\u8fbc\u3093\u3060\u5f8c\u306b\u30ec\u30a4\u30e4\u30fc\u5909\u63db\u3092\u5b9f\u88c5\u3057\u307e\u3059\u3002</p>\n<p>\u73fe\u5728\u3001\u9069\u7528\u3055\u308c\u308b\u306e\u306f int8 \u91cf\u5b50\u5316\u306e\u307f\u3067\u3059\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u6e96\u5099\u3059\u308b\u30ec\u30a4\u30e4\u30fc\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>int8 \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3059\u308b\u304b\u3069\u3046\u304b\u3092\u6307\u5b9a\u3057\u307e\u3059</li>\n<li><span translate=no>_^_2_^_</span>\u30e2\u30c7\u30eb\u306e\u30c7\u30d0\u30a4\u30b9\u3067\u3059</li>\n<li><span translate=no>_^_3_^_</span><span translate=no>_^_4_^_</span>\u5916\u308c\u5024\u306e\u7279\u5fb4\u3092\u5206\u96e2\u3059\u308b\u305f\u3081\u306e\u95be\u5024\u3067\u3059</li>\n<p><em>\u6e96\u5099\u3057\u305f\u30ec\u30a4\u30e4\u30fc\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<p> <span translate=no>_^_0_^_</span> </p>\n": "<p><span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p> Code to load the checkpoint</p>\n": "<p>\u30c1\u30a7\u30c3\u30af\u30dd\u30a4\u30f3\u30c8\u3092\u30ed\u30fc\u30c9\u3059\u308b\u30b3\u30fc\u30c9</p>\n",
|
||||
"<p> Readout layer</p>\n": "<p>\u8aad\u307f\u51fa\u3057\u5c64</p>\n",
|
||||
"<p><a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a> </p>\n": "<p><a href=\"https://github.com/HazyResearch/flash-attention\">\u30d5\u30e9\u30c3\u30b7\u30e5\u30fb\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3</a></p>\n",
|
||||
"<p><span translate=no>_^_0_^_</span> </p>\n": "<p><span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Add RoPE embeddings </p>\n": "<p>RoPe \u57cb\u3081\u8fbc\u307f\u3092\u8ffd\u52a0</p>\n",
|
||||
"<p>Add head dimension </p>\n": "<p>\u982d\u90e8\u5bf8\u6cd5\u3092\u8ffd\u52a0</p>\n",
|
||||
"<p>Add them and the residual connection </p>\n": "<p>\u305d\u308c\u3089\u3068\u6b8b\u308a\u306e\u63a5\u7d9a\u3092\u8ffd\u52a0\u3057\u307e\u3059</p>\n",
|
||||
"<p>Apply mask </p>\n": "<p>\u30de\u30b9\u30af\u3092\u9069\u7528</p>\n",
|
||||
"<p>Attention layer </p>\n": "<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>Attention output transform </p>\n": "<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u51fa\u529b\u5909\u63db</p>\n",
|
||||
"<p>Attention query, key and value transform </p>\n": "<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30af\u30a8\u30ea\u3001\u30ad\u30fc\u3001\u5024\u306e\u5909\u63db</p>\n",
|
||||
"<p>Attention scaling factor </p>\n": "<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u30d5\u30a1\u30af\u30bf\u30fc</p>\n",
|
||||
"<p>Attention softmax </p>\n": "<p>\u6ce8\u610f\u30bd\u30d5\u30c8\u30de\u30c3\u30af\u30b9</p>\n",
|
||||
"<p>Attention softmax module </p>\n": "<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30bd\u30d5\u30c8\u30de\u30c3\u30af\u30b9\u30e2\u30b8\u30e5\u30fc\u30eb</p>\n",
|
||||
"<p>Base for <span translate=no>_^_0_^_</span> </p>\n": "<p>\u306e\u30d9\u30fc\u30b9 <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Cache <span translate=no>_^_0_^_</span> and <span translate=no>_^_1_^_</span> </p>\n": "<p><span translate=no>_^_0_^_</span>\u30ad\u30e3\u30c3\u30b7\u30e5\u3068 <span translate=no>_^_1_^_</span></p>\n",
|
||||
"<p>Cache them </p>\n": "<p>\u305d\u308c\u3089\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b</p>\n",
|
||||
"<p>Calculate <span translate=no>_^_0_^_</span> and <span translate=no>_^_1_^_</span> in fp32 </p>\n": "<p><span translate=no>_^_0_^_</span>\u8a08\u7b97\u3057\u3066 <span translate=no>_^_1_^_</span> fp32 \u3067</p>\n",
|
||||
"<p>Concatenate so that for row <span translate=no>_^_0_^_</span> we have</p>\n<p><span translate=no>_^_1_^_</span> </p>\n": "<p>\u884c\u304c\u6b21\u306e\u3088\u3046\u306b\u306a\u308b\u3088\u3046\u306b\u9023\u7d50\u3057\u307e\u3059 <span translate=no>_^_0_^_</span></p>\n<p><span translate=no>_^_1_^_</span></p>\n",
|
||||
"<p>Concatenate the past </p>\n": "<p>\u904e\u53bb\u3092\u9023\u7d50\u3059\u308b</p>\n",
|
||||
"<p>Concatenate with features that didn't get RoPE embeddings </p>\n": "<p>RoPe \u57cb\u3081\u8fbc\u307f\u306b\u5bfe\u5fdc\u3057\u3066\u3044\u306a\u304b\u3063\u305f\u6a5f\u80fd\u3068\u306e\u9023\u643a</p>\n",
|
||||
"<p>Contraction linear layer </p>\n": "<p>\u53ce\u7e2e\u7dda\u72b6\u5c64</p>\n",
|
||||
"<p>Convert the linear layers </p>\n": "<p>\u7dda\u5f62\u30ec\u30a4\u30e4\u30fc\u306e\u5909\u63db</p>\n",
|
||||
"<p>Convert to fp32 if the current dtype is fp16 </p>\n": "<p>\u73fe\u5728\u306e dtype \u304c fp16 \u306e\u5834\u5408\u306f fp32 \u306b\u5909\u63db</p>\n",
|
||||
"<p>Create mask </p>\n": "<p>\u30de\u30b9\u30af\u4f5c\u6210</p>\n",
|
||||
"<p>Disable auto-casting to fp16 for attention computation </p>\n": "<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u8a08\u7b97\u306e fp16 \u3078\u306e\u81ea\u52d5\u30ad\u30e3\u30b9\u30c8\u3092\u7121\u52b9\u306b\u3059\u308b</p>\n",
|
||||
"<p>Do not cast for bfloat </p>\n": "<p>bfloat\u306b\u306f\u30ad\u30e3\u30b9\u30c8\u3057\u306a\u3044\u3067\u304f\u3060\u3055\u3044</p>\n",
|
||||
"<p>Embedding layer </p>\n": "<p>\u57cb\u3081\u8fbc\u307f\u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>Expansion linear layer </p>\n": "<p>\u62e1\u5f35\u30ea\u30cb\u30a2\u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>FFN first transform </p>\n": "<p>FFN \u30d5\u30a1\u30fc\u30b9\u30c8\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30e0</p>\n",
|
||||
"<p>FFN layer </p>\n": "<p>FFN \u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>FFN second transform </p>\n": "<p>FFN 2 \u756a\u76ee\u306e\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30e0</p>\n",
|
||||
"<p>Final linear layer </p>\n": "<p>\u6700\u7d42\u7dda\u5f62\u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>Final normalization layer </p>\n": "<p>\u6700\u7d42\u6b63\u898f\u5316\u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>GELU activation </p>\n": "<p>GELU \u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3</p>\n",
|
||||
"<p>Get attention weighted values </p>\n": "<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u52a0\u91cd\u5024\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get causal mask </p>\n": "<p>\u30ab\u30b8\u30e5\u30a2\u30eb\u30de\u30b9\u30af\u3092\u30b2\u30c3\u30c8</p>\n",
|
||||
"<p>Get default values if not specified </p>\n": "<p>\u6307\u5b9a\u3057\u306a\u3044\u5834\u5408\u306f\u30c7\u30d5\u30a9\u30eb\u30c8\u5024\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get position indexes <span translate=no>_^_0_^_</span> </p>\n": "<p>\u4f4d\u7f6e\u30a4\u30f3\u30c7\u30c3\u30af\u30b9\u3092\u53d6\u5f97 <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Get query, key and value embeddings (all concatenated). The last dimension size will change from n_hidden -> <span translate=no>_^_0_^_</span> </p>\n": "<p>\u30af\u30a8\u30ea\u3001\u30ad\u30fc\u3001\u5024\u306e\u57cb\u3081\u8fbc\u307f (\u3059\u3079\u3066\u9023\u7d50) \u3092\u53d6\u5f97\u3057\u307e\u3059\u3002\u6700\u5f8c\u306e\u30c7\u30a3\u30e1\u30f3\u30b7\u30e7\u30f3\u30b5\u30a4\u30ba\u304c n_hidden \u304b\u3089\u5909\u66f4\u3055\u308c\u307e\u3059</p>-> <span translate=no>_^_0_^_</span>\n",
|
||||
"<p>Get the actual sequence length </p>\n": "<p>\u5b9f\u969b\u306e\u30b7\u30fc\u30b1\u30f3\u30b9\u9577\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get the past keys and values. These will have shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u904e\u53bb\u306e\u30ad\u30fc\u3068\u5024\u3092\u53d6\u5f97\u3057\u307e\u3059\u3002\u3053\u308c\u3089\u306f\u5f62\u306b\u306a\u308a\u307e\u3059 <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Get the sin and cos values from the cache </p>\n": "<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u304b\u3089 sin \u3068 cos \u306e\u5024\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get the state id's. We use to retrieve previous states and store the next states </p>\n": "<p>\u30b9\u30c6\u30fc\u30c8 ID \u3092\u53d6\u5f97\u3057\u307e\u3059\u3002\u524d\u306e\u30b9\u30c6\u30fc\u30c8\u3092\u53d6\u5f97\u3057\u305f\u308a\u3001\u6b21\u306e\u30b9\u30c6\u30fc\u30c8\u3092\u4fdd\u5b58\u3057\u305f\u308a\u3059\u308b\u306e\u306b\u4f7f\u3044\u307e\u3059\u3002</p>\n",
|
||||
"<p>If there's cache </p>\n": "<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u304c\u3042\u308b\u5834\u5408</p>\n",
|
||||
"<p>If we are caching the states of previous tokens </p>\n": "<p>\u4ee5\u524d\u306e\u30c8\u30fc\u30af\u30f3\u306e\u72b6\u614b\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b\u5834\u5408</p>\n",
|
||||
"<p>Initialize <span translate=no>_^_0_^_</span> </p>\n": "<p>[\u521d\u671f\u5316] <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Initialize <span translate=no>_^_0_^_</span> and <span translate=no>_^_1_^_</span> cache </p>\n": "<p><span translate=no>_^_0_^_</span>\u521d\u671f\u5316\u3068\u30ad\u30e3\u30c3\u30b7\u30e5 <span translate=no>_^_1_^_</span></p>\n",
|
||||
"<p>Layer norm before FFN </p>\n": "<p>FFN \u524d\u306e\u30ec\u30a4\u30e4\u30fc\u30ce\u30eb\u30e0</p>\n",
|
||||
"<p>Layer norm before attention </p>\n": "<p>\u6ce8\u76ee\u3055\u308c\u308b\u524d\u306e\u30ec\u30a4\u30e4\u30fc\u30ce\u30eb\u30e0</p>\n",
|
||||
"<p>Layer normalization before FFN </p>\n": "<p>FFN \u524d\u306e\u30ec\u30a4\u30e4\u30fc\u6b63\u898f\u5316</p>\n",
|
||||
"<p>Layer normalization before attention </p>\n": "<p>\u6ce8\u610f\u524d\u306e\u30ec\u30a4\u30e4\u30fc\u6b63\u898f\u5316</p>\n",
|
||||
"<p>Linear layer for query, key and value </p>\n": "<p>\u30af\u30a8\u30ea\u3001\u30ad\u30fc\u3001\u5024\u306e\u7dda\u5f62\u30ec\u30a4\u30e4\u30fc</p>\n",
|
||||
"<p>NeoX runs attention and feedforward network in parallel </p>\n": "<p>NeoX\u306f\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u3068\u30d5\u30a3\u30fc\u30c9\u30d5\u30a9\u30ef\u30fc\u30c9\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u3092\u4e26\u884c\u3057\u3066\u5b9f\u884c\u3057\u307e\u3059</p>\n",
|
||||
"<p>No cache - simply add RoPE embeddings </p>\n": "<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u306a\u3057-RoPE \u57cb\u3081\u8fbc\u307f\u3092\u8ffd\u52a0\u3059\u308b\u3060\u3051</p>\n",
|
||||
"<p>Number of features for RoPE </p>\n": "<p>RoPE \u306e\u6a5f\u80fd\u306e\u6570</p>\n",
|
||||
"<p>Number of features per head </p>\n": "<p>\u30d8\u30c3\u30c9\u3042\u305f\u308a\u306e\u6a5f\u80fd\u6570</p>\n",
|
||||
"<p>Offset of the current embeddings </p>\n": "<p>\u73fe\u5728\u306e\u57cb\u3081\u8fbc\u307f\u306e\u30aa\u30d5\u30bb\u30c3\u30c8</p>\n",
|
||||
"<p>Only convert the linear layers in the transformer layers </p>\n": "<p>\u30c8\u30e9\u30f3\u30b9\u30ec\u30a4\u30e4\u30fc\u306e\u7dda\u5f62\u30ec\u30a4\u30e4\u30fc\u306e\u307f\u3092\u5909\u63db\u3057\u307e\u3059</p>\n",
|
||||
"<p>Otherwise, use normal attention </p>\n": "<p>\u305d\u308c\u4ee5\u5916\u306e\u5834\u5408\u306f\u3001\u901a\u5e38\u306e\u6ce8\u610f\u3092\u6255\u3063\u3066\u304f\u3060\u3055\u3044</p>\n",
|
||||
"<p>Query and key lengths </p>\n": "<p>\u30af\u30a8\u30ea\u3068\u30ad\u30fc\u306e\u9577\u3055</p>\n",
|
||||
"<p>Readout layer </p>\n": "<p>\u8aad\u307f\u51fa\u3057\u5c64</p>\n",
|
||||
"<p>Reshape from <span translate=no>_^_0_^_</span><a href=\"batch_size, seq_len, n_hidden\">batch_size, seq_len, n_hidden</a>` </p>\n": "<p><span translate=no>_^_0_^_</span><a href=\"batch_size, seq_len, n_hidden\">\u30d0\u30c3\u30c1\u30b5\u30a4\u30ba\u3001\u30b7\u30fc\u30b1\u30f3\u30b9\u756a\u53f7\u3001n_hidden `</a>\u304b\u3089\u5f62\u72b6\u3092\u5909\u66f4</p>\n",
|
||||
"<p>Residual connection </p>\n": "<p>\u6b8b\u7559\u63a5\u7d9a</p>\n",
|
||||
"<p>Return from cache </p>\n": "<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u304b\u3089\u623b\u308b</p>\n",
|
||||
"<p>RoPE embedding module </p>\n": "<p>RoPE \u57cb\u3081\u8fbc\u307f\u30e2\u30b8\u30e5\u30fc\u30eb</p>\n",
|
||||
"<p>RoPE embeddings</p>\n<span translate=no>_^_0_^_</span><p>for <span translate=no>_^_1_^_</span> </p>\n": "<p>\u30ed\u30fc\u30d7\u57cb\u3081\u8fbc\u307f</p>\n<span translate=no>_^_0_^_</span><p>\u306b\u3068\u3063\u3066 <span translate=no>_^_1_^_</span></p>\n",
|
||||
"<p>Save the current state </p>\n": "<p>\u73fe\u5728\u306e\u72b6\u614b\u3092\u4fdd\u5b58\u3059\u308b</p>\n",
|
||||
"<p>Scale attention </p>\n": "<p>\u30b9\u30b1\u30fc\u30eb\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3</p>\n",
|
||||
"<p>Skip if not using int8 quantization </p>\n": "<p>int8 \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u306a\u3044\u5834\u5408\u306f\u30b9\u30ad\u30c3\u30d7</p>\n",
|
||||
"<p>Split into heads by changing the shape to <span translate=no>_^_0_^_</span> </p>\n": "<p>\u5f62\u72b6\u3092\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u5909\u66f4\u3057\u3066\u982d\u90e8\u306b\u5206\u5272\u3057\u307e\u3059 <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Split into query, key and value each of shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u5f62\u72b6\u3054\u3068\u306b\u30af\u30a8\u30ea\u3001\u30ad\u30fc\u3001\u5024\u306b\u5206\u5272 <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Split the features. We apply RoPE to only <span translate=no>_^_0_^_</span> features </p>\n": "<p>\u6a5f\u80fd\u3092\u5206\u5272\u3057\u3066\u304f\u3060\u3055\u3044\u3002RoPE <span translate=no>_^_0_^_</span> \u306f\u6a5f\u80fd\u306b\u306e\u307f\u9069\u7528\u3055\u308c\u307e\u3059</p>\n",
|
||||
"<p>Stack them into shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u305d\u308c\u3089\u3092\u7a4d\u307f\u91cd\u306d\u3066\u5f62\u3092\u6574\u3048\u308b <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>The output is of shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u51fa\u529b\u306f\u6574\u5f62\u3057\u3066\u3044\u307e\u3059 <span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>To cache causal mask </p>\n": "<p>\u56e0\u679c\u30de\u30b9\u30af\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b\u306b\u306f</p>\n",
|
||||
"<p>To store <span translate=no>_^_0_^_</span> for the features </p>\n": "<p><span translate=no>_^_0_^_</span>\u6a5f\u80fd\u7528\u306b\u4fdd\u5b58\u3059\u308b\u306b\u306f</p>\n",
|
||||
"<p>Transformer layer </p>\n": "<p>\u5909\u5727\u5668\u5c64</p>\n",
|
||||
"<p>Transformer layers </p>\n": "<p>\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u5c64</p>\n",
|
||||
"<p>Use <span translate=no>_^_0_^_</span> defined in <a href=\"./utils/llm_int8.html\">utilities</a>. </p>\n": "<p><span translate=no>_^_0_^_</span><a href=\"./utils/llm_int8.html\">\u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3\u3067\u5b9a\u7fa9\u3055\u308c\u3066\u3044\u308b\u7528\u9014</a>\u3002</p>\n",
|
||||
"<p>Use flash attention </p>\n": "<p>\u30d5\u30e9\u30c3\u30b7\u30e5\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u3092\u4f7f\u3046</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> are the embeddings of shape <span translate=no>_^_1_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u5f62\u304c\u57cb\u3081\u8fbc\u307e\u308c\u3066\u3044\u308b\u3082\u306e\u3067\u3059 <span translate=no>_^_1_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> are the token ids of shape <span translate=no>_^_1_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u5f62\u72b6\u306e\u30c8\u30fc\u30af\u30f3ID\u3067\u3059 <span translate=no>_^_1_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> has shape <span translate=no>_^_1_^_</span> </li>\n<li><span translate=no>_^_2_^_</span> is the starting position of <span translate=no>_^_3_^_</span>. This is <span translate=no>_^_4_^_</span> when we have cached the keys and queries of previous positions</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u5f62\u304c\u3042\u308b <span translate=no>_^_1_^_</span></li>\n<li><span translate=no>_^_2_^_</span><span translate=no>_^_3_^_</span>\u306e\u958b\u59cb\u4f4d\u7f6e\u3067\u3059\u3002\u3053\u308c\u306f\u3001<span translate=no>_^_4_^_</span>\u4ee5\u524d\u306e\u30dd\u30b8\u30b7\u30e7\u30f3\u306e\u30ad\u30fc\u3068\u30af\u30a8\u30ea\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3057\u305f\u3068\u304d\u3067\u3059</li></ul>\u3002\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> has shape <span translate=no>_^_1_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u5f62\u304c\u3042\u308b <span translate=no>_^_1_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the embedding size </li>\n<li><span translate=no>_^_1_^_</span> is the number of heads </li>\n<li><span translate=no>_^_2_^_</span> specifies whether to use <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n<p><em>Out implementation doesn't include dropout</em>.</p>\n": "<ul><li><span translate=no>_^_0_^_</span>\u306f\u57cb\u3081\u8fbc\u307f\u30b5\u30a4\u30ba</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u982d\u306e\u6570\u3067\u3059</li>\n<li><span translate=no>_^_2_^_</span><a href=\"https://github.com/HazyResearch/flash-attention\">\u30d5\u30e9\u30c3\u30b7\u30e5\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3059\u308b\u304b\u3069\u3046\u304b\u3092\u6307\u5b9a\u3057\u307e\u3059</a></li></ul>\n<p><em>\u30a2\u30a6\u30c8\u306e\u5b9f\u88c5\u306b\u306f\u30c9\u30ed\u30c3\u30d7\u30a2\u30a6\u30c8\u306f\u542b\u307e\u308c\u3066\u3044\u307e\u305b\u3093</em>\u3002</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the embedding size </li>\n<li><span translate=no>_^_1_^_</span> is the size of the vocabulary</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u306f\u57cb\u3081\u8fbc\u307f\u30b5\u30a4\u30ba</li>\n<li><span translate=no>_^_1_^_</span>\u30dc\u30ad\u30e3\u30d6\u30e9\u30ea\u30fc\u306e\u5927\u304d\u3055\u3067\u3059</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the embedding size</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u306f\u57cb\u3081\u8fbc\u307f\u30b5\u30a4\u30ba</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the number of features for RoPE embeddings </li>\n<li><span translate=no>_^_1_^_</span> is the base for <span translate=no>_^_2_^_</span>, which defaults to <span translate=no>_^_3_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>RoPE \u57cb\u3081\u8fbc\u307f\u306e\u6a5f\u80fd\u306e\u6570\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u304c\u306e\u57fa\u5e95\u3067<span translate=no>_^_2_^_</span>\u3001\u30c7\u30d5\u30a9\u30eb\u30c8\u306f <span translate=no>_^_3_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the size of the vocabulary </li>\n<li><span translate=no>_^_1_^_</span> is the size of the embeddings</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u30dc\u30ad\u30e3\u30d6\u30e9\u30ea\u30fc\u306e\u5927\u304d\u3055\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u57cb\u3081\u8fbc\u307f\u306e\u30b5\u30a4\u30ba\u3067\u3059</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> the number of features in embeddings </li>\n<li><span translate=no>_^_1_^_</span> the number of attention heads </li>\n<li><span translate=no>_^_2_^_</span> percentage of features to add RoPE embeddings </li>\n<li><span translate=no>_^_3_^_</span> masking fill value for attention matrix </li>\n<li><span translate=no>_^_4_^_</span> specifies whether to use <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u57cb\u3081\u8fbc\u307f\u306b\u542b\u307e\u308c\u308b\u6a5f\u80fd\u306e\u6570</li>\n<li><span translate=no>_^_1_^_</span>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30fb\u30d8\u30c3\u30c9\u306e\u6570</li>\n<li><span translate=no>_^_2_^_</span>RoPe \u57cb\u3081\u8fbc\u307f\u3092\u8ffd\u52a0\u3059\u308b\u6a5f\u80fd\u306e\u5272\u5408</li>\n<li><span translate=no>_^_3_^_</span>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30fb\u30de\u30c8\u30ea\u30c3\u30af\u30b9\u306e\u30de\u30b9\u30ad\u30f3\u30b0\u30fb\u30d5\u30a3\u30eb\u5024</li>\n<li><span translate=no>_^_4_^_</span><a href=\"https://github.com/HazyResearch/flash-attention\">\u30d5\u30e9\u30c3\u30b7\u30e5\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3059\u308b\u304b\u3069\u3046\u304b\u3092\u6307\u5b9a\u3057\u307e\u3059</a></li></ul>\n",
|
||||
"GPT-NeoX Model Definition": "GPT-\u30cd\u30aa\u30c3\u30af\u30b9\u30e2\u30c7\u30eb\u5b9a\u7fa9",
|
||||
"This is the model definition of GPT-NeoX.": "\u3053\u308c\u304cGPT-Neox\u306e\u30e2\u30c7\u30eb\u5b9a\u7fa9\u3067\u3059\u3002"
|
||||
}
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,113 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Model</h1>\n<p>Here is the code for layers of GPT-NeoX model and the code to load 20B checkpoint.</p>\n<p>The method <span translate=no>_^_0_^_</span> in the layers load the checkpoints of that layer. The checkpoint loading helpers are on <a href=\"checkpoint.html\"><span translate=no>_^_1_^_</span></a></p>\n": "<h1>GPT-NEOX \u578b\u53f7</h1>\n<p>\u4ee5\u4e0b\u662f GPT-NEOX \u6a21\u578b\u5c42\u7684\u4ee3\u7801\u548c\u52a0\u8f7d 20B \u68c0\u67e5\u70b9\u7684\u4ee3\u7801\u3002</p>\n<p>\u56fe\u5c42<span translate=no>_^_0_^_</span>\u4e2d\u7684\u65b9\u6cd5\u52a0\u8f7d\u8be5\u5c42\u7684\u68c0\u67e5\u70b9\u3002\u68c0\u67e5\u70b9\u52a0\u8f7d\u52a9\u624b\u5df2\u542f\u7528 <a href=\"checkpoint.html\"><span translate=no>_^_1_^_</span></a></p>\n",
|
||||
"<h2>Attention layer</h2>\n": "<h2>\u6ce8\u610f\u5c42</h2>\n",
|
||||
"<h2>Embedding layer</h2>\n<p>This is a standard embeddings layer with code to load the checkpoint.</p>\n": "<h2>\u5d4c\u5165\u5c42</h2>\n<p>\u8fd9\u662f\u4e00\u4e2a\u6807\u51c6\u7684\u5d4c\u5165\u5c42\uff0c\u5176\u4e2d\u5305\u542b\u7528\u4e8e\u52a0\u8f7d\u68c0\u67e5\u70b9\u7684\u4ee3\u7801\u3002</p>\n",
|
||||
"<h2>Feedforward Network</h2>\n": "<h2>\u524d\u9988\u7f51\u7edc</h2>\n",
|
||||
"<h2>Final normalization layer</h2>\n": "<h2>\u6700\u7ec8\u5f52\u4e00\u5316\u5c42</h2>\n",
|
||||
"<h2>Rotary Positional Embeddings</h2>\n<p>GPT-NeoX uses <a href=\"https://arxiv.org/abs/2104.09864\">rotary positional embeddings (RoPE)</a>.</p>\n<p>WE have annotated implementation of RoPE <a href=\"https://nn.labml.ai/transformers/rope/index.html\">here</a> with more notes the theory.</p>\n": "<h2>\u65cb\u8f6c\u4f4d\u7f6e\u5d4c\u5165</h2>\n<p>GPT-NEOX \u4f7f\u7528<a href=\"https://arxiv.org/abs/2104.09864\">\u65cb\u8f6c\u4f4d\u7f6e\u5d4c\u5165\uff08RoP\uff09</a>\u3002</p>\n<p>\u6211\u4eec<a href=\"https://nn.labml.ai/transformers/rope/index.html\">\u5728\u8fd9\u91cc</a>\u6ce8\u91ca\u4e86 RoPe \u7684\u5b9e\u73b0\uff0c\u5e76\u9644\u4e0a\u4e86\u66f4\u591a\u5173\u4e8e\u7406\u8bba\u7684\u6ce8\u91ca\u3002</p>\n",
|
||||
"<h2>Transformer Layer</h2>\n": "<h2>\u53d8\u538b\u5668\u5c42</h2>\n",
|
||||
"<h3>Generator to create layers</h3>\n<p>The layers are generated in the same order as checkpoints.</p>\n<p>It gives <span translate=no>_^_0_^_</span> when a layer is not available; we use the layer indices as NeoX and there are two transformation layers we don't need in our implementation.</p>\n<ul><li><span translate=no>_^_1_^_</span> is the number of tokens in the vocabulary </li>\n<li><span translate=no>_^_2_^_</span> is the number of features in the embeddings </li>\n<li><span translate=no>_^_3_^_</span> is the number of transformer layers </li>\n<li><span translate=no>_^_4_^_</span> is the number of attention heads </li>\n<li><span translate=no>_^_5_^_</span> are the set of layers to be used. All layers will be used if None. This is used to test smaller versions of the model with fewer layers </li>\n<li><span translate=no>_^_6_^_</span> specifies whether to clone the transformer layers (a bit faster) </li>\n<li><span translate=no>_^_7_^_</span> is the data type of the model </li>\n<li><span translate=no>_^_8_^_</span> is the device of the model </li>\n<li><span translate=no>_^_9_^_</span> specifies whether to use int8 quantization </li>\n<li><span translate=no>_^_10_^_</span> is the threshold <span translate=no>_^_11_^_</span> used to separate outlier features </li>\n<li><span translate=no>_^_12_^_</span> specifies whether to use <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n": "<h3>\u7528\u4e8e\u521b\u5efa\u56fe\u5c42\u7684\u751f\u6210\u5668</h3>\n<p>\u56fe\u5c42\u7684\u751f\u6210\u987a\u5e8f\u4e0e\u68c0\u67e5\u70b9\u7684\u751f\u6210\u987a\u5e8f\u76f8\u540c\u3002</p>\n<p>\u5b83\u5728\u56fe\u5c42\u4e0d\u53ef\u7528<span translate=no>_^_0_^_</span>\u65f6\u7ed9\u51fa\uff1b\u6211\u4eec\u5c06\u56fe\u5c42\u7d22\u5f15\u7528\u4f5c NeoX\uff0c\u5e76\u4e14\u5728\u5b9e\u73b0\u4e2d\u4e0d\u9700\u8981\u4e24\u4e2a\u8f6c\u6362\u5c42\u3002</p>\n<ul><li><span translate=no>_^_1_^_</span>\u662f\u8bcd\u6c47\u8868\u4e2d\u7684\u4ee3\u5e01\u6570\u91cf</li>\n<li><span translate=no>_^_2_^_</span>\u662f\u5d4c\u5165\u4e2d\u7684\u7279\u5f81\u6570\u91cf</li>\n<li><span translate=no>_^_3_^_</span>\u662f\u53d8\u538b\u5668\u5c42\u6570</li>\n<li><span translate=no>_^_4_^_</span>\u662f\u6ce8\u610f\u529b\u5934\u7684\u6570\u91cf</li>\n<li><span translate=no>_^_5_^_</span>\u662f\u8981\u4f7f\u7528\u7684\u56fe\u5c42\u96c6\u3002\u5982\u679c\u6ca1\u6709\uff0c\u5219\u5c06\u4f7f\u7528\u6240\u6709\u56fe\u5c42\u3002\u8fd9\u7528\u4e8e\u6d4b\u8bd5\u5c42\u6570\u8f83\u5c11\u7684\u6a21\u578b\u7684\u8f83\u5c0f\u7248\u672c</li>\n<li><span translate=no>_^_6_^_</span>\u6307\u5b9a\u662f\u5426\u514b\u9686\u53d8\u538b\u5668\u5c42\uff08\u5feb\u4e00\u70b9\uff09</li>\n<li><span translate=no>_^_7_^_</span>\u662f\u6a21\u578b\u7684\u6570\u636e\u7c7b\u578b</li>\n<li><span translate=no>_^_8_^_</span>\u662f\u6a21\u578b\u7684\u8bbe\u5907</li>\n<li><span translate=no>_^_9_^_</span>\u6307\u5b9a\u662f\u5426\u4f7f\u7528 int8 \u91cf\u5316</li>\n<li><span translate=no>_^_10_^_</span>\u662f<span translate=no>_^_11_^_</span>\u7528\u4e8e\u5206\u79bb\u5f02\u5e38\u7279\u5f81\u7684\u9608\u503c</li>\n<li><span translate=no>_^_12_^_</span>\u6307\u5b9a\u662f\u5426\u4f7f\u7528 <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n",
|
||||
"<h3>Generator to get layers</h3>\n": "<h3>\u83b7\u53d6\u56fe\u5c42\u7684\u751f\u6210\u5668</h3>\n",
|
||||
"<h3>Generator to load layers</h3>\n": "<h3>\u7528\u4e8e\u52a0\u8f7d\u5c42\u7684\u751f\u6210\u5668</h3>\n",
|
||||
"<h3>Returns the total number of layers</h3>\n": "<h3>\u8fd4\u56de\u603b\u5c42\u6570</h3>\n",
|
||||
"<h3>Rotate the features</h3>\n<p><span translate=no>_^_0_^_</span></p>\n": "<h3>\u65cb\u8f6c\u8981\u7d20</h3>\n<p><span translate=no>_^_0_^_</span></p>\n",
|
||||
"<h4>Calculate the causal mask</h4>\n<ul><li><span translate=no>_^_0_^_</span> has shape <a href=\"batch_size, query_seq_len, key_seq_len, n_heads\">batch_size, query_seq_len, key_seq_len, n_heads</a></li></ul>\n": "<h4>\u8ba1\u7b97\u56e0\u679c\u63a9\u7801</h4>\n<ul><li><span translate=no>_^_0_^_</span>\u6709\u5f62\u72b6 <a href=\"batch_size, query_seq_len, key_seq_len, n_heads\">batch_size\u3001query_seq_len\u3001key_seq_len\u3001n_Heads</a></li></ul>\n",
|
||||
"<h4>Creates and caches a layer</h4>\n<p>Copying cached layers is faster than initializing new layers because it takes time to initialize parameters.</p>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the layer </li>\n<li><span translate=no>_^_1_^_</span> is the function to create the layer </li>\n<p><em>Returns</em> the created layer or a copy of the cached layer</p></ul>\n": "<h4>\u521b\u5efa\u548c\u7f13\u5b58\u56fe\u5c42</h4>\n<p>\u590d\u5236\u7f13\u5b58\u56fe\u5c42\u6bd4\u521d\u59cb\u5316\u65b0\u56fe\u5c42\u8981\u5feb\uff0c\u56e0\u4e3a\u521d\u59cb\u5316\u53c2\u6570\u9700\u8981\u65f6\u95f4\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u5c42\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u521b\u5efa\u56fe\u5c42\u7684\u51fd\u6570</li>\n<p><em>\u8fd4\u56de</em>\u521b\u5efa\u7684\u56fe\u5c42\u6216\u7f13\u5b58\u56fe\u5c42\u7684\u526f\u672c</p></ul>\n",
|
||||
"<h4>Prepares the layer for usage</h4>\n<p>We move the layer to the device and convert it to the correct data type</p>\n<ul><li><span translate=no>_^_0_^_</span> is the layer to prepare </li>\n<p><em>Returns</em> the prepared layer</p></ul>\n": "<h4>\u51c6\u5907\u56fe\u5c42\u4ee5\u4f9b\u4f7f\u7528</h4>\n<p>\u6211\u4eec\u5c06\u56fe\u5c42\u79fb\u52a8\u5230\u8bbe\u5907\u5e76\u5c06\u5176\u8f6c\u6362\u4e3a\u6b63\u786e\u7684\u6570\u636e\u7c7b\u578b</p>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u8981\u51c6\u5907\u7684\u56fe\u5c42</li>\n<p><em>\u8fd4\u56de</em>\u51c6\u5907\u597d\u7684\u56fe\u5c42</p></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p> <a id=\"post_load_prepare\"></a></p>\n<h3>Layer transformations after loading the checkpoint</h3>\n<p>This function implements layer transformations after loading the checkpoint.</p>\n<p>Currently, it only applies the int8 quantization.</p>\n<ul><li><span translate=no>_^_0_^_</span> is the layer to prepare </li>\n<li><span translate=no>_^_1_^_</span> specifies whether to use int8 quantization </li>\n<li><span translate=no>_^_2_^_</span> is the device of the model </li>\n<li><span translate=no>_^_3_^_</span> is the threshold <span translate=no>_^_4_^_</span> used to separate outlier features </li>\n<p><em>Returns</em> the prepared layer</p></ul>\n": "<p><a id=\"post_load_prepare\"></a></p>\n<h3>\u52a0\u8f7d\u68c0\u67e5\u70b9\u540e\u7684\u56fe\u5c42\u53d8\u6362</h3>\n<p>\u6b64\u51fd\u6570\u5728\u52a0\u8f7d\u68c0\u67e5\u70b9\u540e\u5b9e\u73b0\u5c42\u8f6c\u6362\u3002</p>\n<p>\u76ee\u524d\uff0c\u5b83\u4ec5\u5e94\u7528 int8 \u91cf\u5316\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u8981\u51c6\u5907\u7684\u56fe\u5c42</li>\n<li><span translate=no>_^_1_^_</span>\u6307\u5b9a\u662f\u5426\u4f7f\u7528 int8 \u91cf\u5316</li>\n<li><span translate=no>_^_2_^_</span>\u662f\u8be5\u578b\u53f7\u7684\u8bbe\u5907</li>\n<li><span translate=no>_^_3_^_</span>\u662f<span translate=no>_^_4_^_</span>\u7528\u4e8e\u5206\u9694\u5f02\u5e38\u503c\u8981\u7d20\u7684\u9608\u503c</li>\n<p><em>\u8fd4\u56de</em>\u51c6\u5907\u597d\u7684\u56fe\u5c42</p></ul>\n",
|
||||
"<p> <span translate=no>_^_0_^_</span> </p>\n": "<p><span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p> Code to load the checkpoint</p>\n": "<p>\u52a0\u8f7d\u68c0\u67e5\u70b9\u7684\u4ee3\u7801</p>\n",
|
||||
"<p> Readout layer</p>\n": "<p>\u8bfb\u51fa\u5c42</p>\n",
|
||||
"<p><a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a> </p>\n": "<p><a href=\"https://github.com/HazyResearch/flash-attention\">\u95ea\u5149\u6ce8\u610f</a></p>\n",
|
||||
"<p><span translate=no>_^_0_^_</span> </p>\n": "<p><span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Add RoPE embeddings </p>\n": "<p>\u6dfb\u52a0\u7ef3\u7d22\u5d4c\u5165</p>\n",
|
||||
"<p>Add head dimension </p>\n": "<p>\u6dfb\u52a0\u5934\u90e8\u5c3a\u5bf8</p>\n",
|
||||
"<p>Add them and the residual connection </p>\n": "<p>\u6dfb\u52a0\u5b83\u4eec\u548c\u5269\u4f59\u7684\u8fde\u63a5</p>\n",
|
||||
"<p>Apply mask </p>\n": "<p>\u6d82\u62b9\u9762\u819c</p>\n",
|
||||
"<p>Attention layer </p>\n": "<p>\u6ce8\u610f\u5c42</p>\n",
|
||||
"<p>Attention output transform </p>\n": "<p>\u6ce8\u610f\u529b\u8f93\u51fa\u53d8\u6362</p>\n",
|
||||
"<p>Attention query, key and value transform </p>\n": "<p>\u6ce8\u610f\u529b\u67e5\u8be2\u3001\u5173\u952e\u548c\u4ef7\u503c\u8f6c\u6362</p>\n",
|
||||
"<p>Attention scaling factor </p>\n": "<p>\u6ce8\u610f\u529b\u7f29\u653e\u7cfb\u6570</p>\n",
|
||||
"<p>Attention softmax </p>\n": "<p>\u6ce8\u610f softmax</p>\n",
|
||||
"<p>Attention softmax module </p>\n": "<p>\u6ce8\u610f softmax \u6a21\u5757</p>\n",
|
||||
"<p>Base for <span translate=no>_^_0_^_</span> </p>\n": "<p>\u57fa\u5730<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Cache <span translate=no>_^_0_^_</span> and <span translate=no>_^_1_^_</span> </p>\n": "<p>\u7f13\u5b58<span translate=no>_^_0_^_</span>\u548c<span translate=no>_^_1_^_</span></p>\n",
|
||||
"<p>Cache them </p>\n": "<p>\u7f13\u5b58\u5b83\u4eec</p>\n",
|
||||
"<p>Calculate <span translate=no>_^_0_^_</span> and <span translate=no>_^_1_^_</span> in fp32 </p>\n": "<p><span translate=no>_^_1_^_</span>\u5728 fp32 \u4e2d\u8ba1\u7b97<span translate=no>_^_0_^_</span>\u548c</p>\n",
|
||||
"<p>Concatenate so that for row <span translate=no>_^_0_^_</span> we have</p>\n<p><span translate=no>_^_1_^_</span> </p>\n": "<p>\u8fde\u63a5\u8fd9\u6837<span translate=no>_^_0_^_</span>\u6211\u4eec\u5c31\u6709 row</p>\n<p><span translate=no>_^_1_^_</span></p>\n",
|
||||
"<p>Concatenate the past </p>\n": "<p>\u4e32\u8054\u8fc7\u53bb</p>\n",
|
||||
"<p>Concatenate with features that didn't get RoPE embeddings </p>\n": "<p>\u8fde\u63a5\u672a\u83b7\u5f97 RoPe \u5d4c\u5165\u7684\u529f\u80fd</p>\n",
|
||||
"<p>Contraction linear layer </p>\n": "<p>\u6536\u7f29\u7ebf\u6027\u5c42</p>\n",
|
||||
"<p>Convert the linear layers </p>\n": "<p>\u8f6c\u6362\u7ebf\u6027\u56fe\u5c42</p>\n",
|
||||
"<p>Convert to fp32 if the current dtype is fp16 </p>\n": "<p>\u5982\u679c\u5f53\u524d\u6570\u636e\u7c7b\u578b\u4e3a fp16\uff0c\u5219\u8f6c\u6362\u4e3a fp32</p>\n",
|
||||
"<p>Create mask </p>\n": "<p>\u521b\u5efa\u906e\u7f69</p>\n",
|
||||
"<p>Disable auto-casting to fp16 for attention computation </p>\n": "<p>\u7981\u7528\u81ea\u52a8\u6295\u5c04\u5230 fp16 \u4ee5\u8fdb\u884c\u6ce8\u610f\u529b\u8ba1\u7b97</p>\n",
|
||||
"<p>Do not cast for bfloat </p>\n": "<p>\u4e0d\u8981\u4e3a bfloat \u8fdb\u884c\u6295\u5c04</p>\n",
|
||||
"<p>Embedding layer </p>\n": "<p>\u5d4c\u5165\u5c42</p>\n",
|
||||
"<p>Expansion linear layer </p>\n": "<p>\u6269\u5c55\u7ebf\u6027\u5c42</p>\n",
|
||||
"<p>FFN first transform </p>\n": "<p>FFN \u9996\u6b21\u6539\u9020</p>\n",
|
||||
"<p>FFN layer </p>\n": "<p>FFN \u5c42</p>\n",
|
||||
"<p>FFN second transform </p>\n": "<p>FFN \u7b2c\u4e8c\u6b21\u53d8\u6362</p>\n",
|
||||
"<p>Final linear layer </p>\n": "<p>\u6700\u540e\u7684\u7ebf\u6027\u5c42</p>\n",
|
||||
"<p>Final normalization layer </p>\n": "<p>\u6700\u7ec8\u5f52\u4e00\u5316\u5c42</p>\n",
|
||||
"<p>GELU activation </p>\n": "<p>GELU \u6fc0\u6d3b</p>\n",
|
||||
"<p>Get attention weighted values </p>\n": "<p>\u83b7\u53d6\u6ce8\u610f\u529b\u52a0\u6743\u503c</p>\n",
|
||||
"<p>Get causal mask </p>\n": "<p>\u83b7\u5f97\u56e0\u679c\u53e3\u7f69</p>\n",
|
||||
"<p>Get default values if not specified </p>\n": "<p>\u5982\u679c\u672a\u6307\u5b9a\uff0c\u5219\u83b7\u53d6\u9ed8\u8ba4\u503c</p>\n",
|
||||
"<p>Get position indexes <span translate=no>_^_0_^_</span> </p>\n": "<p>\u83b7\u53d6\u5934\u5bf8\u6307\u6570<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Get query, key and value embeddings (all concatenated). The last dimension size will change from n_hidden -> <span translate=no>_^_0_^_</span> </p>\n": "<p>\u83b7\u53d6\u67e5\u8be2\u3001\u952e\u548c\u503c\u5d4c\u5165\uff08\u5168\u90e8\u4e32\u8054\uff09\u3002\u6700\u540e\u4e00\u4e2a\u7ef4\u5ea6\u5927\u5c0f\u5c06\u4ece n_hidden \u66f4\u6539\u4e3a-><span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Get the actual sequence length </p>\n": "<p>\u83b7\u53d6\u5b9e\u9645\u5e8f\u5217\u957f\u5ea6</p>\n",
|
||||
"<p>Get the past keys and values. These will have shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u83b7\u53d6\u8fc7\u53bb\u7684\u952e\u548c\u503c\u3002\u8fd9\u4e9b\u4f1a\u6709\u5f62\u72b6<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Get the sin and cos values from the cache </p>\n": "<p>\u4ece\u7f13\u5b58\u4e2d\u83b7\u53d6 sin \u548c cos \u503c</p>\n",
|
||||
"<p>Get the state id's. We use to retrieve previous states and store the next states </p>\n": "<p>\u83b7\u53d6\u72b6\u6001 ID\u3002\u6211\u4eec\u7528\u5b83\u6765\u68c0\u7d22\u4ee5\u524d\u7684\u72b6\u6001\u5e76\u5b58\u50a8\u4e0b\u4e00\u4e2a\u72b6\u6001</p>\n",
|
||||
"<p>If there's cache </p>\n": "<p>\u5982\u679c\u6709\u7f13\u5b58</p>\n",
|
||||
"<p>If we are caching the states of previous tokens </p>\n": "<p>\u5982\u679c\u6211\u4eec\u6b63\u5728\u7f13\u5b58\u4e4b\u524d\u4ee4\u724c\u7684\u72b6\u6001</p>\n",
|
||||
"<p>Initialize <span translate=no>_^_0_^_</span> </p>\n": "<p>\u521d\u59cb\u5316<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Initialize <span translate=no>_^_0_^_</span> and <span translate=no>_^_1_^_</span> cache </p>\n": "<p>\u521d\u59cb\u5316<span translate=no>_^_0_^_</span>\u5e76<span translate=no>_^_1_^_</span>\u7f13\u5b58</p>\n",
|
||||
"<p>Layer norm before FFN </p>\n": "<p>FFN \u4e4b\u524d\u7684\u5206\u5c42\u89c4\u8303</p>\n",
|
||||
"<p>Layer norm before attention </p>\n": "<p>\u6ce8\u610f\u4e4b\u524d\u5148\u8fdb\u884c\u5206\u5c42\u89c4\u8303</p>\n",
|
||||
"<p>Layer normalization before FFN </p>\n": "<p>FFN \u4e4b\u524d\u7684\u5c42\u6807\u51c6\u5316</p>\n",
|
||||
"<p>Layer normalization before attention </p>\n": "<p>\u6ce8\u610f\u4e4b\u524d\u7684\u56fe\u5c42\u89c4\u8303\u5316</p>\n",
|
||||
"<p>Linear layer for query, key and value </p>\n": "<p>\u7528\u4e8e\u67e5\u8be2\u3001\u952e\u548c\u503c\u7684\u7ebf\u6027\u56fe\u5c42</p>\n",
|
||||
"<p>NeoX runs attention and feedforward network in parallel </p>\n": "<p>NeoX \u5e76\u884c\u8fd0\u884c\u6ce8\u610f\u529b\u548c\u524d\u9988\u7f51\u7edc</p>\n",
|
||||
"<p>No cache - simply add RoPE embeddings </p>\n": "<p>\u6ca1\u6709\u7f13\u5b58-\u53ea\u9700\u6dfb\u52a0 RoPe \u5d4c\u5165\u5373\u53ef</p>\n",
|
||||
"<p>Number of features for RoPE </p>\n": "<p>ROPE \u7684\u8981\u7d20\u6570\u91cf</p>\n",
|
||||
"<p>Number of features per head </p>\n": "<p>\u6bcf\u5934\u7279\u5f81\u6570</p>\n",
|
||||
"<p>Offset of the current embeddings </p>\n": "<p>\u5f53\u524d\u5d4c\u5165\u7684\u504f\u79fb\u91cf</p>\n",
|
||||
"<p>Only convert the linear layers in the transformer layers </p>\n": "<p>\u4ec5\u8f6c\u6362\u53d8\u538b\u5668\u5c42\u4e2d\u7684\u7ebf\u6027\u5c42</p>\n",
|
||||
"<p>Otherwise, use normal attention </p>\n": "<p>\u5426\u5219\uff0c\u8bf7\u6b63\u5e38\u6ce8\u610f</p>\n",
|
||||
"<p>Query and key lengths </p>\n": "<p>\u67e5\u8be2\u548c\u5bc6\u94a5\u957f\u5ea6</p>\n",
|
||||
"<p>Readout layer </p>\n": "<p>\u8bfb\u51fa\u5c42</p>\n",
|
||||
"<p>Reshape from <span translate=no>_^_0_^_</span><a href=\"batch_size, seq_len, n_hidden\">batch_size, seq_len, n_hidden</a>` </p>\n": "<p>\u4ece<span translate=no>_^_0_^_</span> <a href=\"batch_size, seq_len, n_hidden\">batch_size\u3001seq_len\u3001n_hidden \u8fdb\u884c\u91cd\u5851</a> `</p>\n",
|
||||
"<p>Residual connection </p>\n": "<p>\u5269\u4f59\u8fde\u63a5</p>\n",
|
||||
"<p>Return from cache </p>\n": "<p>\u4ece\u7f13\u5b58\u4e2d\u8fd4\u56de</p>\n",
|
||||
"<p>RoPE embedding module </p>\n": "<p>\u7ef3\u7d22\u5d4c\u5165\u6a21\u5757</p>\n",
|
||||
"<p>RoPE embeddings</p>\n<span translate=no>_^_0_^_</span><p>for <span translate=no>_^_1_^_</span> </p>\n": "<p>\u7ef3\u7d22\u5d4c\u5165</p>\n<span translate=no>_^_0_^_</span><p>\u5bf9\u4e8e<span translate=no>_^_1_^_</span></p>\n",
|
||||
"<p>Save the current state </p>\n": "<p>\u4fdd\u5b58\u5f53\u524d\u72b6\u6001</p>\n",
|
||||
"<p>Scale attention </p>\n": "<p>\u7f29\u653e\u6ce8\u610f\u529b</p>\n",
|
||||
"<p>Skip if not using int8 quantization </p>\n": "<p>\u5982\u679c\u4e0d\u4f7f\u7528 int8 \u91cf\u5316\u5219\u8df3\u8fc7</p>\n",
|
||||
"<p>Split into heads by changing the shape to <span translate=no>_^_0_^_</span> </p>\n": "<p>\u901a\u8fc7\u5c06\u5f62\u72b6\u6539\u4e3a\u5206\u6210\u5934\u90e8<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Split into query, key and value each of shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u5206\u4e3a\u67e5\u8be2\u3001\u952e\u548c\u503c\u5404\u5f62\u72b6<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Split the features. We apply RoPE to only <span translate=no>_^_0_^_</span> features </p>\n": "<p>\u62c6\u5206\u8981\u7d20\u3002\u6211\u4eec\u4ec5\u5c06 RoPe \u5e94\u7528\u4e8e\u8981<span translate=no>_^_0_^_</span>\u7d20</p>\n",
|
||||
"<p>Stack them into shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u5c06\u5b83\u4eec\u5806\u53e0\u6210\u5f62\u72b6<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>The output is of shape <span translate=no>_^_0_^_</span> </p>\n": "<p>\u8f93\u51fa\u7684\u5f62\u72b6\u662f\u8fd9\u6837\u7684<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>To cache causal mask </p>\n": "<p>\u7f13\u5b58\u56e0\u679c\u63a9\u7801</p>\n",
|
||||
"<p>To store <span translate=no>_^_0_^_</span> for the features </p>\n": "<p>\u4e3a\u8981\u7d20\u5b58\u50a8<span translate=no>_^_0_^_</span></p>\n",
|
||||
"<p>Transformer layer </p>\n": "<p>\u53d8\u538b\u5668\u5c42</p>\n",
|
||||
"<p>Transformer layers </p>\n": "<p>\u53d8\u538b\u5668\u5c42</p>\n",
|
||||
"<p>Use <span translate=no>_^_0_^_</span> defined in <a href=\"./utils/llm_int8.html\">utilities</a>. </p>\n": "<p>\u4f7f\u7528\u5728<a href=\"./utils/llm_int8.html\">\u5b9e\u7528\u7a0b\u5e8f</a>\u4e2d<span translate=no>_^_0_^_</span>\u5b9a\u4e49\u3002</p>\n",
|
||||
"<p>Use flash attention </p>\n": "<p>\u4f7f\u7528\u95ea\u5149\u706f\u6ce8\u610f\u529b</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> are the embeddings of shape <span translate=no>_^_1_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u5f62\u72b6\u7684\u5d4c\u5165<span translate=no>_^_1_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> are the token ids of shape <span translate=no>_^_1_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u5f62\u72b6\u7684\u4ee4\u724c ID<span translate=no>_^_1_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> has shape <span translate=no>_^_1_^_</span> </li>\n<li><span translate=no>_^_2_^_</span> is the starting position of <span translate=no>_^_3_^_</span>. This is <span translate=no>_^_4_^_</span> when we have cached the keys and queries of previous positions</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u6709\u5f62\u72b6<span translate=no>_^_1_^_</span></li>\n<li><span translate=no>_^_2_^_</span>\u662f\u7684\u8d77\u59cb\u4f4d\u7f6e<span translate=no>_^_3_^_</span>\u3002\u8fd9\u662f\u6211\u4eec\u7f13\u5b58\u5148\u524d\u4f4d\u7f6e\u7684\u952e\u548c\u67e5\u8be2<span translate=no>_^_4_^_</span>\u7684\u65f6\u5019</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> has shape <span translate=no>_^_1_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u6709\u5f62\u72b6<span translate=no>_^_1_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the embedding size </li>\n<li><span translate=no>_^_1_^_</span> is the number of heads </li>\n<li><span translate=no>_^_2_^_</span> specifies whether to use <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n<p><em>Out implementation doesn't include dropout</em>.</p>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u5d4c\u5165\u5927\u5c0f</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u5934\u6570</li>\n<li><span translate=no>_^_2_^_</span>\u6307\u5b9a\u662f\u5426\u4f7f\u7528 <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n<p><em>Out \u7684\u5b9e\u73b0\u4e0d\u5305\u62ec\u9000\u51fa</em>\u3002</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the embedding size </li>\n<li><span translate=no>_^_1_^_</span> is the size of the vocabulary</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u5d4c\u5165\u7684\u5927\u5c0f</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u8bcd\u6c47\u91cf\u7684\u5927\u5c0f</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the embedding size</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u5d4c\u5165\u7684\u5927\u5c0f</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the number of features for RoPE embeddings </li>\n<li><span translate=no>_^_1_^_</span> is the base for <span translate=no>_^_2_^_</span>, which defaults to <span translate=no>_^_3_^_</span></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f RoPe \u5d4c\u5165\u7684\u8981\u7d20\u6570\u91cf</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u7684\u57fa\u7840<span translate=no>_^_2_^_</span>\uff0c\u9ed8\u8ba4\u4e3a<span translate=no>_^_3_^_</span></li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the size of the vocabulary </li>\n<li><span translate=no>_^_1_^_</span> is the size of the embeddings</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u8bcd\u6c47\u91cf\u7684\u5927\u5c0f</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u5d4c\u5165\u7684\u5927\u5c0f</li></ul>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> the number of features in embeddings </li>\n<li><span translate=no>_^_1_^_</span> the number of attention heads </li>\n<li><span translate=no>_^_2_^_</span> percentage of features to add RoPE embeddings </li>\n<li><span translate=no>_^_3_^_</span> masking fill value for attention matrix </li>\n<li><span translate=no>_^_4_^_</span> specifies whether to use <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u5d4c\u5165\u4e2d\u7684\u7279\u5f81\u6570\u91cf</li>\n<li><span translate=no>_^_1_^_</span>\u6ce8\u610f\u529b\u5934\u7684\u6570\u91cf</li>\n<li><span translate=no>_^_2_^_</span>\u6dfb\u52a0 RoPE \u5d4c\u5165\u7684\u529f\u80fd\u767e\u5206\u6bd4</li>\n<li><span translate=no>_^_3_^_</span>\u63a9\u76d6\u6ce8\u610f\u529b\u77e9\u9635\u7684\u586b\u5145\u503c</li>\n<li><span translate=no>_^_4_^_</span>\u6307\u5b9a\u662f\u5426\u4f7f\u7528 <a href=\"https://github.com/HazyResearch/flash-attention\">FlashAttention</a></li></ul>\n",
|
||||
"GPT-NeoX Model Definition": "GPT-NEOX \u578b\u53f7\u5b9a\u4e49",
|
||||
"This is the model definition of GPT-NeoX.": "\u8fd9\u662f GPT-NEOX \u7684\u6a21\u578b\u5b9a\u4e49\u3002"
|
||||
}
|
||||
@@ -0,0 +1,3 @@
|
||||
{
|
||||
"GPT-NeoX": "GPT \u30cd\u30aa\u30c3\u30af\u30b9"
|
||||
}
|
||||
@@ -0,0 +1,3 @@
|
||||
{
|
||||
"GPT-NeoX": "\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca"
|
||||
}
|
||||
@@ -0,0 +1,3 @@
|
||||
{
|
||||
"GPT-NeoX": "GPT-neox"
|
||||
}
|
||||
@@ -0,0 +1,5 @@
|
||||
{
|
||||
"<h1>Samples</h1>\n<ul><li><a href=\"generate.html\">Generating text</a> </li>\n<li><a href=\"finetune.html\">Fine tuning the biases with pipeline-parallel training</a></li></ul>\n": "<h1>[\u30b5\u30f3\u30d7\u30eb]</h1>\n<ul><li><a href=\"generate.html\">\u30c6\u30ad\u30b9\u30c8\u306e\u751f\u6210</a></li>\n<li><a href=\"finetune.html\">\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u4e26\u5217\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u306b\u3088\u308b\u30d0\u30a4\u30a2\u30b9\u306e\u5fae\u8abf\u6574</a></li></ul>\n",
|
||||
"Samples": "[\u30b5\u30f3\u30d7\u30eb]",
|
||||
"Samples for inference and fine-tuning": "\u63a8\u8ad6\u3068\u5fae\u8abf\u6574\u7528\u306e\u30b5\u30f3\u30d7\u30eb"
|
||||
}
|
||||
@@ -0,0 +1,5 @@
|
||||
{
|
||||
"<h1>Samples</h1>\n<ul><li><a href=\"generate.html\">Generating text</a> </li>\n<li><a href=\"finetune.html\">Fine tuning the biases with pipeline-parallel training</a></li></ul>\n": "<h1>\u0dc3\u0dcf\u0db8\u0dca\u0db4\u0dbd</h1>\n<ul><li><a href=\"generate.html\">\u0db4\u0dd9\u0dc5 \u0d8b\u0dad\u0dca\u0db4\u0dcf\u0daf\u0db1\u0dba</a> </li>\n<li><a href=\"finetune.html\">\u0db1\u0dbd \u0db8\u0dcf\u0dbb\u0dca\u0d9c-\u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb \u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4\u0dc0 \u0dc3\u0db8\u0d9f \u0d85\u0d9c\u0dad\u0dd3\u0db1\u0dca \u0dc4\u0ddc\u0db3\u0dd2\u0db1\u0dca \u0dc3\u0dd4\u0dc3\u0dbb \u0d9a\u0dd2\u0dbb\u0dd3\u0db8</a></li></ul>\n",
|
||||
"Samples": "\u0dc3\u0dcf\u0db8\u0dca\u0db4\u0dbd",
|
||||
"Samples for inference and fine-tuning": "\u0d85\u0db1\u0dd4\u0db8\u0dcf\u0db1\u0dba \u0dc3\u0dc4 \u0db8\u0db1\u0dcf\u0dc0 \u0dc3\u0dd4\u0dc3\u0dbb \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0dc3\u0dcf\u0db8\u0dca\u0db4\u0dbd"
|
||||
}
|
||||
@@ -0,0 +1,5 @@
|
||||
{
|
||||
"<h1>Samples</h1>\n<ul><li><a href=\"generate.html\">Generating text</a> </li>\n<li><a href=\"finetune.html\">Fine tuning the biases with pipeline-parallel training</a></li></ul>\n": "<h1>\u6837\u54c1</h1>\n<ul><li><a href=\"generate.html\">\u751f\u6210\u6587\u672c</a></li>\n</ul><li><a href=\"finetune.html\">\u901a\u8fc7\u7ba1\u9053\u5e73\u884c\u8bad\u7ec3\u5fae\u8c03\u504f\u5dee</a></li>\n",
|
||||
"Samples": "\u6837\u54c1",
|
||||
"Samples for inference and fine-tuning": "\u7528\u4e8e\u63a8\u7406\u548c\u5fae\u8c03\u7684\u6837\u672c"
|
||||
}
|
||||
@@ -0,0 +1,21 @@
|
||||
{
|
||||
"<h1>Fine Tune GPT-NeoX</h1>\n<p>This shows how to fine tune GPT-NeoX with pipeline parallelism.</p>\n": "<h1>\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30f3 GPT-\u30cd\u30aa\u30c3\u30af\u30b9</h1>\n<p>\u3053\u308c\u306f\u3001\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u4e26\u5217\u51e6\u7406\u3067GPT-Neox\u3092\u5fae\u8abf\u6574\u3059\u308b\u65b9\u6cd5\u3092\u793a\u3057\u3066\u3044\u307e\u3059\u3002</p>\n",
|
||||
"<h3>Create fine tuner for biases</h3>\n": "<h3>\u30d0\u30a4\u30a2\u30b9\u306e\u5fae\u8abf\u6574\u5668\u306e\u4f5c\u6210</h3>\n",
|
||||
"<h3>Create pipeline parallel model</h3>\n": "<h3>\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u4e26\u5217\u30e2\u30c7\u30eb\u306e\u4f5c\u6210</h3>\n",
|
||||
"<h3>Load GPT-NeoX layers</h3>\n": "<h3>GPT-Neox \u30ec\u30a4\u30e4\u30fc\u3092\u30ed\u30fc\u30c9</h3>\n",
|
||||
"<h4>Tiny Shakespeare dataset</h4>\n": "<h4>\u5c0f\u3055\u306a\u30b7\u30a7\u30a4\u30af\u30b9\u30d4\u30a2\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8</h4>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create Fairscale Pipe module </p>\n": "<p>\u30d5\u30a7\u30a2\u30b9\u30b1\u30fc\u30eb\u30d1\u30a4\u30d7\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f5c\u6210</p>\n",
|
||||
"<p>Create experiment </p>\n": "<p>\u5b9f\u9a13\u3092\u4f5c\u6210</p>\n",
|
||||
"<p>Create the Pipe module </p>\n": "<p>\u30d1\u30a4\u30d7\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f5c\u6210\u3059\u308b</p>\n",
|
||||
"<p>Devices for each GPU </p>\n": "<p>\u5404 GPU \u306e\u30c7\u30d0\u30a4\u30b9</p>\n",
|
||||
"<p>Get the layer distribution across GPUs </p>\n": "<p>GPU \u5168\u4f53\u306e\u30ec\u30a4\u30e4\u30fc\u5206\u5e03\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Initialize configs </p>\n": "<p>\u30b3\u30f3\u30d5\u30a3\u30b0\u3092\u521d\u671f\u5316</p>\n",
|
||||
"<p>Initialize the model. Do this before the loop for cleaner logs. </p>\n": "<p>\u30e2\u30c7\u30eb\u3092\u521d\u671f\u5316\u3057\u307e\u3059\u3002\u3053\u308c\u3092\u30eb\u30fc\u30d7\u306e\u524d\u306b\u884c\u3063\u3066\u3001\u30ed\u30b0\u3092\u30af\u30ea\u30fc\u30f3\u30a2\u30c3\u30d7\u3057\u3066\u304f\u3060\u3055\u3044\u3002</p>\n",
|
||||
"<p>Make sure the finetuner is initialized </p>\n": "<p>\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30ca\u30fc\u304c\u521d\u671f\u5316\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u3066\u304f\u3060\u3055\u3044</p>\n",
|
||||
"<p>Mark biases as trainable </p>\n": "<p>\u30d0\u30a4\u30a2\u30b9\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u53ef\u80fd\u3068\u30de\u30fc\u30af\u3059\u308b</p>\n",
|
||||
"<p>Start the experiment </p>\n": "<p>\u5b9f\u9a13\u3092\u59cb\u3081\u308b</p>\n",
|
||||
"<p>Train </p>\n": "<p>\u5217\u8eca</p>\n",
|
||||
"Fine Tune GPT-NeoX": "\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30f3 GPT-\u30cd\u30aa\u30c3\u30af\u30b9",
|
||||
"Fine tune GPT-NeoX biases with Fairscale pipeline parallel module": "\u30d5\u30a7\u30a2\u30b9\u30b1\u30fc\u30eb\u30fb\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u30fb\u30d1\u30e9\u30ec\u30eb\u30fb\u30e2\u30b8\u30e5\u30fc\u30eb\u306b\u3088\u308bGPT-Neox\u30d0\u30a4\u30a2\u30b9\u306e\u5fae\u8abf\u6574"
|
||||
}
|
||||
@@ -0,0 +1,21 @@
|
||||
{
|
||||
"<h1>Fine Tune GPT-NeoX</h1>\n<p>This shows how to fine tune GPT-NeoX with pipeline parallelism.</p>\n": "<h1>\u0dc3\u0dd2\u0dc4\u0dd2\u0db1\u0dca\u0da7\u0dd2\u0dba\u0dd4\u0db1\u0dca \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca</h1>\n<p>\u0db1\u0dbd\u0db8\u0dcf\u0dbb\u0dca\u0d9c \u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb\u0d9a\u0dbb\u0dab\u0dba \u0dc3\u0db8\u0d9f \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dc4\u0ddc\u0db3\u0dd2\u0db1\u0dca \u0dc3\u0dd4\u0dc3\u0dbb \u0d9a\u0dbb\u0db1\u0dca\u0db1\u0dda \u0d9a\u0dd9\u0dc3\u0dda\u0daf\u0dd0\u0dba\u0dd2 \u0db8\u0dd9\u0dba\u0dd2\u0db1\u0dca \u0db4\u0dd9\u0db1\u0dca\u0dc0\u0dba\u0dd2. </p>\n",
|
||||
"<h3>Create fine tuner for biases</h3>\n": "<h3>\u0d85\u0d9c\u0dad\u0dd3\u0db1\u0dca\u0dc3\u0db3\u0dc4\u0dcf \u0dc3\u0dd2\u0dc4\u0dd2\u0db1\u0dca \u0dc3\u0dd4\u0dc3\u0dbb\u0d9a\u0dba\u0d9a\u0dca \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1</h3>\n",
|
||||
"<h3>Create pipeline parallel model</h3>\n": "<h3>\u0db1\u0dbd\u0db8\u0dcf\u0dbb\u0dca\u0d9c \u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0d9a\u0dca \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1</h3>\n",
|
||||
"<h3>Load GPT-NeoX layers</h3>\n": "<h3>\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca\u0dc3\u0dca\u0dae\u0dbb \u0db4\u0da7\u0dc0\u0db1\u0dca\u0db1</h3>\n",
|
||||
"<h4>Tiny Shakespeare dataset</h4>\n": "<h4>\u0d9a\u0dd4\u0da9\u0dcf\u0dc2\u0dda\u0d9a\u0dca\u0dc3\u0dca\u0db4\u0dd2\u0dba\u0dbb\u0dca \u0daf\u0dad\u0dca\u0dad \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba</h4>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Create Fairscale Pipe module </p>\n": "<p>\u0dc3\u0dcf\u0db0\u0dcf\u0dbb\u0dab\u0db1\u0dbd \u0db8\u0ddc\u0da9\u0dd2\u0dba\u0dd4\u0dbd\u0dba \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Create experiment </p>\n": "<p>\u0d85\u0dad\u0dca\u0dc4\u0daf\u0dcf\u0db6\u0dd0\u0dbd\u0dd3\u0db8 \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Create the Pipe module </p>\n": "<p>\u0db4\u0dba\u0dd2\u0db4\u0dca\u0db4\u0db8\u0ddc\u0da9\u0dd2\u0dba\u0dd4\u0dbd\u0dba \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Devices for each GPU </p>\n": "<p>\u0d91\u0d9a\u0dca\u0d91\u0d9a\u0dca GPU \u0dc3\u0db3\u0dc4\u0dcf \u0d8b\u0db4\u0dcf\u0d82\u0d9c </p>\n",
|
||||
"<p>Get the layer distribution across GPUs </p>\n": "<p>GPU\u0dc4\u0dbb\u0dc4\u0dcf \u0dc3\u0dca\u0dae\u0dbb \u0dc0\u0dca\u0dba\u0dcf\u0db4\u0dca\u0dad\u0dd2\u0dba \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Initialize configs </p>\n": "<p>\u0dc0\u0dd2\u0db1\u0dca\u0dba\u0dcf\u0dc3\u0d86\u0dbb\u0db8\u0dca\u0db7 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Initialize the model. Do this before the loop for cleaner logs. </p>\n": "<p>\u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0d86\u0dbb\u0db8\u0dca\u0db7 \u0d9a\u0dbb\u0db1\u0dca\u0db1. \u0db4\u0dd2\u0dbb\u0dd2\u0dc3\u0dd2\u0daf\u0dd4 \u0dbd\u0ddc\u0d9c\u0dca \u0dc3\u0db3\u0dc4\u0dcf \u0dbd\u0dd6\u0db4\u0dba\u0da7 \u0db4\u0dd9\u0dbb \u0db8\u0dd9\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1. </p>\n",
|
||||
"<p>Make sure the finetuner is initialized </p>\n": "<p>\u0db8\u0dd9\u0db8finetuner \u0d86\u0dbb\u0db8\u0dca\u0db7 \u0d9a\u0dbb \u0d87\u0dad\u0dd2 \u0db6\u0dc0\u0da7 \u0dc0\u0d9c \u0db6\u0dbd\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Mark biases as trainable </p>\n": "<p>\u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4\u0d9a\u0dc5 \u0dc4\u0dd0\u0d9a\u0dd2 \u0dbd\u0dd9\u0dc3 \u0d85\u0d9c\u0dad\u0dd3\u0db1\u0dca \u0dc3\u0dbd\u0d9a\u0dd4\u0dab\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Start the experiment </p>\n": "<p>\u0d85\u0dad\u0dca\u0dc4\u0daf\u0dcf\u0db6\u0dd0\u0dbd\u0dd3\u0db8 \u0d86\u0dbb\u0db8\u0dca\u0db7 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Train </p>\n": "<p>\u0daf\u0dd4\u0db8\u0dca\u0dbb\u0dd2\u0dba </p>\n",
|
||||
"Fine Tune GPT-NeoX": "\u0dc3\u0dd2\u0dc4\u0dd2\u0db1\u0dca \u0da7\u0dd2\u0dba\u0dd4\u0db1\u0dca \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca",
|
||||
"Fine tune GPT-NeoX biases with Fairscale pipeline parallel module": "\u0dc6\u0dd9\u0dba\u0dcf\u0dbb\u0dca\u0dc3\u0dca\u0d9a\u0dda\u0dbd\u0dca \u0db1\u0dbd \u0db8\u0dcf\u0dbb\u0dca\u0d9c \u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb \u0db8\u0ddc\u0da9\u0dd2\u0dba\u0dd4\u0dbd\u0dba \u0dc3\u0db8\u0d9f \u0dc4\u0ddc\u0db3 \u0dc3\u0dd4\u0dc3\u0dbb \u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0d85\u0d9c\u0dad\u0dd3\u0db1\u0dca"
|
||||
}
|
||||
@@ -0,0 +1,21 @@
|
||||
{
|
||||
"<h1>Fine Tune GPT-NeoX</h1>\n<p>This shows how to fine tune GPT-NeoX with pipeline parallelism.</p>\n": "<h1>Fine Tune GPT-NEOX</h1>\n<p>\u8fd9\u8bf4\u660e\u4e86\u5982\u4f55\u5229\u7528\u7ba1\u9053\u5e76\u884c\u5ea6\u5bf9 GPT-NEOX \u8fdb\u884c\u5fae\u8c03\u3002</p>\n",
|
||||
"<h3>Create fine tuner for biases</h3>\n": "<h3>\u4e3a\u504f\u89c1\u521b\u5efa\u7cbe\u7ec6\u7684\u8c03\u8c10\u5668</h3>\n",
|
||||
"<h3>Create pipeline parallel model</h3>\n": "<h3>\u521b\u5efa\u7ba1\u9053\u5e76\u884c\u6a21\u578b</h3>\n",
|
||||
"<h3>Load GPT-NeoX layers</h3>\n": "<h3>\u52a0\u8f7d GPT-NEOX \u56fe\u5c42</h3>\n",
|
||||
"<h4>Tiny Shakespeare dataset</h4>\n": "<h4>\u5c0f\u838e\u58eb\u6bd4\u4e9a\u6570\u636e\u96c6</h4>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create Fairscale Pipe module </p>\n": "<p>\u521b\u5efa\u516c\u5e73\u89c4\u6a21\u7ba1\u9053\u6a21\u5757</p>\n",
|
||||
"<p>Create experiment </p>\n": "<p>\u521b\u5efa\u5b9e\u9a8c</p>\n",
|
||||
"<p>Create the Pipe module </p>\n": "<p>\u521b\u5efa\u7ba1\u9053\u6a21\u5757</p>\n",
|
||||
"<p>Devices for each GPU </p>\n": "<p>\u6bcf\u4e2a GPU \u7684\u8bbe\u5907</p>\n",
|
||||
"<p>Get the layer distribution across GPUs </p>\n": "<p>\u83b7\u53d6\u8de8 GPU \u7684\u5c42\u5206\u5e03</p>\n",
|
||||
"<p>Initialize configs </p>\n": "<p>\u521d\u59cb\u5316\u914d\u7f6e</p>\n",
|
||||
"<p>Initialize the model. Do this before the loop for cleaner logs. </p>\n": "<p>\u521d\u59cb\u5316\u6a21\u578b\u3002\u5728\u5faa\u73af\u4e4b\u524d\u6267\u884c\u6b64\u64cd\u4f5c\u4ee5\u83b7\u5f97\u66f4\u6e05\u6670\u7684\u65e5\u5fd7\u3002</p>\n",
|
||||
"<p>Make sure the finetuner is initialized </p>\n": "<p>\u786e\u4fdd\u5fae\u8c03\u5668\u5df2\u521d\u59cb\u5316</p>\n",
|
||||
"<p>Mark biases as trainable </p>\n": "<p>\u5c06\u504f\u89c1\u6807\u8bb0\u4e3a\u53ef\u8bad\u7ec3</p>\n",
|
||||
"<p>Start the experiment </p>\n": "<p>\u5f00\u59cb\u5b9e\u9a8c</p>\n",
|
||||
"<p>Train </p>\n": "<p>\u706b\u8f66</p>\n",
|
||||
"Fine Tune GPT-NeoX": "Fine Tune GPT-NEOX",
|
||||
"Fine tune GPT-NeoX biases with Fairscale pipeline parallel module": "\u4f7f\u7528 Fairscale \u7ba1\u9053\u5e76\u884c\u6a21\u5757\u5fae\u8c03 GPT-NEOX \u504f\u5dee"
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"<h1>Generate Text with GPT-NeoX</h1>\n<p>This shows how to generate text from GPT-NeoX with a single GPU.</p>\n<p>This needs a GPU with more than 45GB memory.</p>\n": "<h1>GPT-\u30cd\u30aa\u30c3\u30af\u30b9\u3067\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210</h1>\n<p>\u3053\u308c\u306f\u3001\u5358\u4e00\u306eGPU\u3067GPT-Neox\u304b\u3089\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210\u3059\u308b\u65b9\u6cd5\u3092\u793a\u3057\u3066\u3044\u307e\u3059\u3002</p>\n<p>\u3053\u308c\u306b\u306f\u300145 GB \u4ee5\u4e0a\u306e\u30e1\u30e2\u30ea\u3092\u642d\u8f09\u3057\u305f GPU \u304c\u5fc5\u8981\u3067\u3059\u3002</p>\n",
|
||||
"<h2>Generate text</h2>\n": "<h2>\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210</h2>\n",
|
||||
"<h3>Predict the next token</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the model </li>\n<li><span translate=no>_^_1_^_</span> are the input token ids </li>\n<li><span translate=no>_^_2_^_</span> is the device of the model</li></ul>\n": "<h3>\u6b21\u306e\u30c8\u30fc\u30af\u30f3\u3092\u4e88\u6e2c</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30e2\u30c7\u30eb\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u5165\u529b\u30c8\u30fc\u30af\u30f3 ID</li>\n<li><span translate=no>_^_2_^_</span>\u30e2\u30c7\u30eb\u306e\u30c7\u30d0\u30a4\u30b9\u3067\u3059</li></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Append the predicted token </p>\n": "<p>\u4e88\u6e2c\u30c8\u30fc\u30af\u30f3\u3092\u8ffd\u52a0</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u7aef\u672b</p>\n",
|
||||
"<p>Eval model </p>\n": "<p>\u8a55\u4fa1\u30e2\u30c7\u30eb</p>\n",
|
||||
"<p>Get next token. Note that we only feed the last token to the model because we cache the key/value pairs of previous tokens. </p>\n": "<p>\u6b21\u306e\u30c8\u30fc\u30af\u30f3\u3092\u5165\u624b\u3057\u3066\u304f\u3060\u3055\u3044\u3002\u4ee5\u524d\u306e\u30c8\u30fc\u30af\u30f3\u306e\u30ad\u30fc\u3068\u5024\u306e\u30da\u30a2\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b\u306e\u3067\u3001\u6700\u5f8c\u306e\u30c8\u30fc\u30af\u30f3\u306e\u307f\u3092\u30e2\u30c7\u30eb\u306b\u30d5\u30a3\u30fc\u30c9\u3059\u308b\u3053\u3068\u306b\u6ce8\u610f\u3057\u3066\u304f\u3060\u3055\u3044</p>\u3002\n",
|
||||
"<p>Get the tokens </p>\n": "<p>\u30c8\u30fc\u30af\u30f3\u3092\u5165\u624b</p>\n",
|
||||
"<p>Get token ids </p>\n": "<p>\u30c8\u30fc\u30af\u30f3 ID \u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Imports </p>\n": "<p>\u8f38\u5165</p>\n",
|
||||
"<p>List of layers to load. This is used for testing. You can assign a subset of layers like <span translate=no>_^_0_^_</span> so that it only loads the first to transformer layers. </p>\n": "<p>\u30ed\u30fc\u30c9\u3059\u308b\u30ec\u30a4\u30e4\u30fc\u306e\u30ea\u30b9\u30c8\u3002\u3053\u308c\u306f\u30c6\u30b9\u30c8\u306b\u4f7f\u7528\u3055\u308c\u307e\u3059\u3002<span translate=no>_^_0_^_</span>\u306e\u3088\u3046\u306b\u30ec\u30a4\u30e4\u30fc\u306e\u30b5\u30d6\u30bb\u30c3\u30c8\u3092\u5272\u308a\u5f53\u3066\u3066\u3001\u6700\u521d\u306e\u30ec\u30a4\u30e4\u30fc\u306e\u307f\u3092\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30ec\u30a4\u30e4\u30fc\u306b\u8aad\u307f\u8fbc\u3080\u3053\u3068\u304c\u3067\u304d\u307e\u3059</p>\u3002\n",
|
||||
"<p>Load layers </p>\n": "<p>\u30ec\u30a4\u30e4\u30fc\u3092\u30ed\u30fc\u30c9</p>\n",
|
||||
"<p>Predict 100 tokens </p>\n": "<p>\u30c8\u30fc\u30af\u30f3\u3092100\u500b\u4e88\u6e2c\u3059\u308b</p>\n",
|
||||
"<p>Print </p>\n": "<p>\u30d7\u30ea\u30f3\u30c8</p>\n",
|
||||
"<p>Prompt to complete </p>\n": "<p>\u5b8c\u4e86\u3092\u4fc3\u3059\u30d7\u30ed\u30f3\u30d7\u30c8</p>\n",
|
||||
"<p>Return predicted token </p>\n": "<p>\u4e88\u6e2c\u30c8\u30fc\u30af\u30f3\u3092\u8fd4\u3059</p>\n",
|
||||
"<p>Run the model </p>\n": "<p>\u30e2\u30c7\u30eb\u3092\u5b9f\u884c</p>\n",
|
||||
"<p>Set the state to use cached activations </p>\n": "<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u3055\u308c\u305f\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3059\u308b\u3088\u3046\u306b\u72b6\u614b\u3092\u8a2d\u5b9a\u3057\u307e\u3059</p>\n",
|
||||
"<p>Setup <a href=\"../utils/cache.html\">cache</a> to cache intermediate key/value pairs for faster generation </p>\n": "<p><a href=\"../utils/cache.html\">\u751f\u6210\u3092\u9ad8\u901f\u5316\u3059\u308b\u305f\u3081\u306b\u4e2d\u9593\u30ad\u30fc\u3068\u5024\u306e\u30da\u30a2\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b\u3088\u3046\u306b\u30ad\u30e3\u30c3\u30b7\u30e5\u3092\u8a2d\u5b9a</a></p>\n",
|
||||
"Generate Text with GPT-NeoX": "GPT-\u30cd\u30aa\u30c3\u30af\u30b9\u3067\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210"
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"<h1>Generate Text with GPT-NeoX</h1>\n<p>This shows how to generate text from GPT-NeoX with a single GPU.</p>\n<p>This needs a GPU with more than 45GB memory.</p>\n": "<h1>GPT-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca\u0dc3\u0db8\u0d9f \u0db4\u0dd9\u0dc5 \u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h1>\n<p>\u0dad\u0db1\u0dd2GPU \u0d91\u0d9a\u0d9a\u0dca \u0dc3\u0db8\u0d9f GPT-neox \u0dc0\u0dd9\u0dad\u0dd2\u0db1\u0dca \u0db4\u0dd9\u0dc5 \u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1\u0dda \u0d9a\u0dd9\u0dc3\u0dda\u0daf \u0dba\u0db1\u0dca\u0db1 \u0db8\u0dd9\u0dba\u0dd2\u0db1\u0dca \u0db4\u0dd9\u0db1\u0dca\u0dc0\u0dba\u0dd2. </p>\n<p>\u0db8\u0dda\u0dc3\u0db3\u0dc4\u0dcf 45GB \u0da7 \u0dc0\u0da9\u0dcf \u0dc0\u0dd0\u0da9\u0dd2 \u0db8\u0dad\u0d9a\u0dba\u0d9a\u0dca \u0dc3\u0dc4\u0dd2\u0dad GPU \u0d91\u0d9a\u0d9a\u0dca \u0d85\u0dc0\u0dc1\u0dca\u0dba \u0dc0\u0dda. </p>\n",
|
||||
"<h2>Generate text</h2>\n": "<h2>\u0db4\u0dd9\u0dc5\u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h2>\n",
|
||||
"<h3>Predict the next token</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the model </li>\n<li><span translate=no>_^_1_^_</span> are the input token ids </li>\n<li><span translate=no>_^_2_^_</span> is the device of the model</li></ul>\n": "<h3>\u0d8a\u0dc5\u0d9f\u0da7\u0ddd\u0d9a\u0db1\u0dba \u0db4\u0dd4\u0dbb\u0ddd\u0d9a\u0dae\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0d86\u0daf\u0dcf\u0db1 \u0da7\u0ddd\u0d9a\u0db1 \u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca \u0dc0\u0dda </li>\n</ul><li><span translate=no>_^_2_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0dda \u0d8b\u0db4\u0dcf\u0d82\u0d9c\u0dba \u0dc0\u0dda</li>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Append the predicted token </p>\n": "<p>\u0db4\u0dd4\u0dbb\u0ddd\u0d9a\u0dae\u0db1\u0dba\u0d9a\u0dc5 \u0da7\u0ddd\u0d9a\u0db1\u0dba \u0d91\u0d9a\u0dca \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Device </p>\n": "<p>\u0d8b\u0db4\u0dcf\u0d82\u0d9c\u0dba </p>\n",
|
||||
"<p>Eval model </p>\n": "<p>\u0d91\u0dc0\u0dcf\u0dbd\u0dca\u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba </p>\n",
|
||||
"<p>Get next token. Note that we only feed the last token to the model because we cache the key/value pairs of previous tokens. </p>\n": "<p>\u0d8a\u0dc5\u0d9f\u0da7\u0ddd\u0d9a\u0db1\u0dba \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1. \u0db4\u0dd9\u0dbb \u0da7\u0ddd\u0d9a\u0db1 \u0dc0\u0dbd \u0dba\u0dad\u0dd4\u0dbb/\u0d85\u0d9c\u0dba \u0dba\u0dd4\u0d9c\u0dbd \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2 \u0d9a\u0dbb\u0db1 \u0db1\u0dd2\u0dc3\u0dcf \u0d85\u0db4\u0dd2 \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0da7 \u0d85\u0dc0\u0dc3\u0dcf\u0db1 \u0da7\u0ddd\u0d9a\u0db1\u0dba \u0db4\u0db8\u0dab\u0d9a\u0dca \u0db4\u0ddd\u0dc2\u0dab\u0dba \u0d9a\u0dbb\u0db1 \u0db6\u0dc0 \u0dc3\u0dbd\u0d9a\u0db1\u0dca\u0db1. </p>\n",
|
||||
"<p>Get the tokens </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Get token ids </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0dca\u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca\u0db4\u0dad\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Imports </p>\n": "<p>\u0d86\u0db1\u0dba\u0db1 </p>\n",
|
||||
"<p>List of layers to load. This is used for testing. You can assign a subset of layers like <span translate=no>_^_0_^_</span> so that it only loads the first to transformer layers. </p>\n": "<p>\u0db4\u0dd0\u0da7\u0dc0\u0dd2\u0dba\u0dba\u0dd4\u0dad\u0dd4 \u0dc3\u0dca\u0dae\u0dbb \u0dbd\u0dd0\u0dba\u0dd2\u0dc3\u0dca\u0dad\u0dd4\u0dc0. \u0db8\u0dd9\u0dba \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0dc0\u0dda. \u0da7\u0dca\u0dbb\u0dcf\u0db1\u0dca\u0dc3\u0dca\u0dc6\u0ddd\u0db8\u0dbb\u0dca \u0dc3\u0dca\u0dae\u0dbb \u0dc0\u0dbd\u0da7 \u0db4\u0dc5\u0db8\u0dd4 \u0db4\u0da7\u0dc0\u0db1\u0dd4 \u0dbd\u0db6\u0db1 <span translate=no>_^_0_^_</span> \u0db4\u0dbb\u0dd2\u0daf\u0dd2 \u0d94\u0db6\u0da7 \u0dc0\u0dd0\u0db1\u0dd2 \u0dc3\u0dca\u0dae\u0dbb \u0d8b\u0db4 \u0d9a\u0dd4\u0dbd\u0d9a\u0dba\u0d9a\u0dca \u0db4\u0dd0\u0dc0\u0dbb\u0dd2\u0dba \u0dc4\u0dd0\u0d9a\u0dd2\u0dba. </p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u0dc3\u0dca\u0dae\u0dbb\u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Predict 100 tokens </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1100 \u0d9a\u0dca \u0db4\u0dd4\u0dbb\u0ddd\u0d9a\u0dae\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Print </p>\n": "<p>\u0db8\u0dd4\u0daf\u0dca\u0dbb\u0dab\u0dba </p>\n",
|
||||
"<p>Prompt to complete </p>\n": "<p>\u0dc3\u0db8\u0dca\u0db4\u0dd6\u0dbb\u0dca\u0dab\u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0da7 \u0dc0\u0dd2\u0db8\u0dc3\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Return predicted token </p>\n": "<p>\u0db4\u0dd4\u0dbb\u0ddd\u0d9a\u0dae\u0db1\u0dba\u0d9a\u0dc5 \u0da7\u0ddd\u0d9a\u0db1\u0dba \u0d86\u0db4\u0dc3\u0dd4 \u0dba\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Run the model </p>\n": "<p>\u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0db0\u0dcf\u0dc0\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Set the state to use cached activations </p>\n": "<p>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dc3\u0d9a\u0dca\u0dbb\u0dd2\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dca \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0da7 \u0dbb\u0dcf\u0da2\u0dca\u0dba\u0dba \u0dc3\u0d9a\u0dc3\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Setup <a href=\"../utils/cache.html\">cache</a> to cache intermediate key/value pairs for faster generation </p>\n": "<p>\u0dc0\u0dda\u0d9c\u0dc0\u0dad\u0dca\u0d8b\u0dad\u0dca\u0db4\u0dcf\u0daf\u0db1\u0dba \u0dc3\u0db3\u0dc4\u0dcf \u0d85\u0dad\u0dbb\u0db8\u0dd0\u0daf\u0dd2 \u0dba\u0dad\u0dd4\u0dbb/\u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0dca \u0dba\u0dd4\u0d9c\u0dbd <a href=\"../utils/cache.html\">\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2</a> \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0da7 \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba \u0db4\u0dd2\u0dc4\u0dd2\u0da7\u0dd4\u0dc0\u0db1\u0dca\u0db1 </p>\n",
|
||||
"Generate Text with GPT-NeoX": "GPT-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dc3\u0db8\u0d9f \u0db4\u0dd9\u0dc5 \u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1"
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"<h1>Generate Text with GPT-NeoX</h1>\n<p>This shows how to generate text from GPT-NeoX with a single GPU.</p>\n<p>This needs a GPU with more than 45GB memory.</p>\n": "<h1>\u4f7f\u7528 GPT-NEOX \u751f\u6210\u6587\u672c</h1>\n<p>\u8fd9\u8bf4\u660e\u4e86\u5982\u4f55\u4f7f\u7528\u5355\u4e2a GPU \u4ece GPT-NEOX \u751f\u6210\u6587\u672c\u3002</p>\n<p>\u8fd9\u9700\u8981\u4e00\u4e2a\u5185\u5b58\u8d85\u8fc745GB\u7684GPU\u3002</p>\n",
|
||||
"<h2>Generate text</h2>\n": "<h2>\u751f\u6210\u6587\u672c</h2>\n",
|
||||
"<h3>Predict the next token</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the model </li>\n<li><span translate=no>_^_1_^_</span> are the input token ids </li>\n<li><span translate=no>_^_2_^_</span> is the device of the model</li></ul>\n": "<h3>\u9884\u6d4b\u4e0b\u4e00\u4e2a\u4ee3\u5e01</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u6a21\u7279\u5417</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u8f93\u5165\u4ee4\u724c ID</li>\n<li><span translate=no>_^_2_^_</span>\u662f\u8be5\u578b\u53f7\u7684\u8bbe\u5907</li></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Append the predicted token </p>\n": "<p>\u8ffd\u52a0\u9884\u6d4b\u7684\u4ee4\u724c</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u8bbe\u5907</p>\n",
|
||||
"<p>Eval model </p>\n": "<p>\u8bc4\u4f30\u6a21\u578b</p>\n",
|
||||
"<p>Get next token. Note that we only feed the last token to the model because we cache the key/value pairs of previous tokens. </p>\n": "<p>\u83b7\u53d6\u4e0b\u4e00\u4e2a\u4ee4\u724c\u3002\u8bf7\u6ce8\u610f\uff0c\u6211\u4eec\u53ea\u5c06\u6700\u540e\u4e00\u4e2a\u4ee4\u724c\u63d0\u4f9b\u7ed9\u6a21\u578b\uff0c\u56e0\u4e3a\u6211\u4eec\u7f13\u5b58\u4e86\u5148\u524d\u4ee4\u724c\u7684\u952e/\u503c\u5bf9\u3002</p>\n",
|
||||
"<p>Get the tokens </p>\n": "<p>\u83b7\u53d6\u4ee3\u5e01</p>\n",
|
||||
"<p>Get token ids </p>\n": "<p>\u83b7\u53d6\u4ee3\u5e01 ID</p>\n",
|
||||
"<p>Imports </p>\n": "<p>\u8fdb\u53e3</p>\n",
|
||||
"<p>List of layers to load. This is used for testing. You can assign a subset of layers like <span translate=no>_^_0_^_</span> so that it only loads the first to transformer layers. </p>\n": "<p>\u8981\u52a0\u8f7d\u7684\u56fe\u5c42\u5217\u8868\u3002\u8fd9\u7528\u4e8e\u6d4b\u8bd5\u3002\u60a8\u53ef\u4ee5\u5c06\u5c42\u7684\u5b50\u96c6\u5206\u914d\u7ed9\u53d8\u538b\u5668\u5c42\uff0c<span translate=no>_^_0_^_</span>\u4f7f\u5176\u4ec5\u5c06\u7b2c\u4e00\u4e2a\u5c42\u52a0\u8f7d\u5230\u53d8\u538b\u5668\u5c42\u3002</p>\n",
|
||||
"<p>Load layers </p>\n": "<p>\u52a0\u8f7d\u56fe\u5c42</p>\n",
|
||||
"<p>Predict 100 tokens </p>\n": "<p>\u9884\u6d4b 100 \u4e2a\u4ee3\u5e01</p>\n",
|
||||
"<p>Print </p>\n": "<p>\u6253\u5370</p>\n",
|
||||
"<p>Prompt to complete </p>\n": "<p>\u63d0\u793a\u5b8c\u6210</p>\n",
|
||||
"<p>Return predicted token </p>\n": "<p>\u8fd4\u56de\u9884\u6d4b\u7684\u4ee3\u5e01</p>\n",
|
||||
"<p>Run the model </p>\n": "<p>\u8fd0\u884c\u6a21\u578b</p>\n",
|
||||
"<p>Set the state to use cached activations </p>\n": "<p>\u8bbe\u7f6e\u72b6\u6001\u4ee5\u4f7f\u7528\u7f13\u5b58\u7684\u6fc0\u6d3b</p>\n",
|
||||
"<p>Setup <a href=\"../utils/cache.html\">cache</a> to cache intermediate key/value pairs for faster generation </p>\n": "<p>\u8bbe\u7f6e<a href=\"../utils/cache.html\">\u7f13\u5b58</a>\u4ee5\u7f13\u5b58\u4e2d\u95f4\u952e/\u503c\u5bf9\u4ee5\u52a0\u5feb\u751f\u6210\u901f\u5ea6</p>\n",
|
||||
"Generate Text with GPT-NeoX": "\u4f7f\u7528 GPT-NEOX \u751f\u6210\u6587\u672c"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h1>Generate Text with GPT-NeoX using LLM.int8() quantization</h1>\n<p>This shows how to generate text from GPT-NeoX using <a href=\"../utils/llm_int8.html\">LLM.int8() quantization</a>.</p>\n<p>This needs a GPU with 24GB memory.</p>\n": "<h1>LLM.int8 () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066 GPT-Neox \u3067\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210</h1>\n<p>\u3053\u308c\u306f\u3001<a href=\"../utils/llm_int8.html\">LLM.int8</a> () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066 GPT-Neox \u304b\u3089\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210\u3059\u308b\u65b9\u6cd5\u3092\u793a\u3057\u3066\u3044\u307e\u3059\u3002</p>\n<p>\u3053\u308c\u306b\u306f 24 GB \u306e\u30e1\u30e2\u30ea\u3092\u642d\u8f09\u3057\u305f GPU \u304c\u5fc5\u8981\u3067\u3059\u3002</p>\n",
|
||||
"<h2>Generate text</h2>\n": "<h2>\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210</h2>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Append the predicted token </p>\n": "<p>\u4e88\u6e2c\u30c8\u30fc\u30af\u30f3\u3092\u8ffd\u52a0</p>\n",
|
||||
"<p>Clear cache and print memory summary for debugging </p>\n": "<p>\u30c7\u30d0\u30c3\u30b0\u7528\u306b\u30ad\u30e3\u30c3\u30b7\u30e5\u3092\u30af\u30ea\u30a2\u3057\u3066\u30e1\u30e2\u30ea\u306e\u6982\u8981\u3092\u5370\u5237</p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p><span translate=no>_^_0_^_</span>\u30e2\u30c7\u30eb\u4f5c\u6210</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u7aef\u672b</p>\n",
|
||||
"<p>Get next token. Note that we only feed the last token to the model because we cache the key/value pairs of previous tokens. </p>\n": "<p>\u6b21\u306e\u30c8\u30fc\u30af\u30f3\u3092\u5165\u624b\u3057\u3066\u304f\u3060\u3055\u3044\u3002\u4ee5\u524d\u306e\u30c8\u30fc\u30af\u30f3\u306e\u30ad\u30fc\u3068\u5024\u306e\u30da\u30a2\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b\u306e\u3067\u3001\u6700\u5f8c\u306e\u30c8\u30fc\u30af\u30f3\u306e\u307f\u3092\u30e2\u30c7\u30eb\u306b\u30d5\u30a3\u30fc\u30c9\u3059\u308b\u3053\u3068\u306b\u6ce8\u610f\u3057\u3066\u304f\u3060\u3055\u3044</p>\u3002\n",
|
||||
"<p>Get token ids </p>\n": "<p>\u30c8\u30fc\u30af\u30f3 ID \u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Load layers in float16 into CPU. We convert the layers to int8 later, because doing that on the fly after loading layers to GPU causes CUDA memory fragmentation (about 3GB memory can get lost due to fragmentation). </p>\n": "<p>float16 \u306e\u30ec\u30a4\u30e4\u30fc\u3092 CPU \u306b\u30ed\u30fc\u30c9\u3057\u307e\u3059\u3002\u30ec\u30a4\u30e4\u30fc\u3092GPU\u306b\u30ed\u30fc\u30c9\u3057\u305f\u5f8c\u306b\u305d\u306e\u5834\u3067\u3053\u308c\u3092\u884c\u3046\u3068\u3001CUDA\u30e1\u30e2\u30ea\u306e\u65ad\u7247\u5316\u304c\u767a\u751f\u3059\u308b\u305f\u3081\u3001\u5f8c\u3067\u30ec\u30a4\u30e4\u30fc\u3092int8\u306b\u5909\u63db\u3057\u307e\u3059\uff08\u30d5\u30e9\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u306b\u3088\u308a\u7d043GB\u306e\u30e1\u30e2\u30ea\u304c\u5931\u308f\u308c\u308b\u53ef\u80fd\u6027\u304c\u3042\u308a\u307e\u3059</p>\uff09\u3002\n",
|
||||
"<p>Predict 100 tokens </p>\n": "<p>\u30c8\u30fc\u30af\u30f3\u3092100\u500b\u4e88\u6e2c\u3059\u308b</p>\n",
|
||||
"<p>Print </p>\n": "<p>\u30d7\u30ea\u30f3\u30c8</p>\n",
|
||||
"<p>Run the model. We use the <a href=\"generate.html\"><span translate=no>_^_0_^_</span></a> function defined in <a href=\"generate.html\"><span translate=no>_^_1_^_</span></a> </p>\n": "<p>\u30e2\u30c7\u30eb\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<a href=\"generate.html\"><span translate=no>_^_0_^_</span></a>\u3067\u5b9a\u7fa9\u3055\u308c\u3066\u3044\u308b\u95a2\u6570\u3092\u4f7f\u7528\u3057\u307e\u3059 <a href=\"generate.html\"><span translate=no>_^_1_^_</span></a></p>\n",
|
||||
"<p>Set the state to use cached activations </p>\n": "<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u3055\u308c\u305f\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3059\u308b\u3088\u3046\u306b\u72b6\u614b\u3092\u8a2d\u5b9a\u3057\u307e\u3059</p>\n",
|
||||
"<p>Setup <a href=\"../utils/cache.html\">cache</a> to cache intermediate key/value pairs for faster generation </p>\n": "<p><a href=\"../utils/cache.html\">\u751f\u6210\u3092\u9ad8\u901f\u5316\u3059\u308b\u305f\u3081\u306b\u4e2d\u9593\u30ad\u30fc\u3068\u5024\u306e\u30da\u30a2\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b\u3088\u3046\u306b\u30ad\u30e3\u30c3\u30b7\u30e5\u3092\u8a2d\u5b9a</a></p>\n",
|
||||
"<p>This reduces CUDA memory fragmentation </p>\n": "<p>\u3053\u308c\u306b\u3088\u308a\u3001CUDA \u30e1\u30e2\u30ea\u306e\u65ad\u7247\u5316\u304c\u6e1b\u5c11\u3057\u307e\u3059\u3002</p>\n",
|
||||
"Generate Text with GPT-NeoX using LLM.int8() quantization": "LLM.int8 () \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066 GPT-Neox \u3067\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h1>Generate Text with GPT-NeoX using LLM.int8() quantization</h1>\n<p>This shows how to generate text from GPT-NeoX using <a href=\"../utils/llm_int8.html\">LLM.int8() quantization</a>.</p>\n<p>This needs a GPU with 24GB memory.</p>\n": "<h1>LLM.INT8() \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca GPT-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dc3\u0db8\u0d9f \u0db4\u0dd9\u0dc5 \u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h1>\n<p><a href=\"../utils/llm_int8.html\">LLM.INT8 () \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba</a>\u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca GPT-Neox \u0dc0\u0dd9\u0dad\u0dd2\u0db1\u0dca \u0db4\u0dd9\u0dc5 \u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1\u0dda \u0d9a\u0dd9\u0dc3\u0dda\u0daf \u0dba\u0db1\u0dca\u0db1 \u0db8\u0dd9\u0dba\u0dd2\u0db1\u0dca \u0db4\u0dd9\u0db1\u0dca\u0dc0\u0dba\u0dd2. </p>\n<p>\u0db8\u0dda\u0dc3\u0db3\u0dc4\u0dcf 24GB \u0db8\u0dad\u0d9a\u0dba\u0d9a\u0dca \u0dc3\u0dc4\u0dd2\u0dad GPU \u0d91\u0d9a\u0d9a\u0dca \u0d85\u0dc0\u0dc1\u0dca\u0dba \u0dc0\u0dda. </p>\n",
|
||||
"<h2>Generate text</h2>\n": "<h2>\u0db4\u0dd9\u0dc5\u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h2>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Append the predicted token </p>\n": "<p>\u0db4\u0dd4\u0dbb\u0ddd\u0d9a\u0dae\u0db1\u0dba\u0d9a\u0dc5 \u0da7\u0ddd\u0d9a\u0db1\u0dba \u0d91\u0d9a\u0dca \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Clear cache and print memory summary for debugging </p>\n": "<p>\u0db1\u0dd2\u0daf\u0ddc\u0dc3\u0dca\u0d9a\u0dbb\u0dab\u0dba\u0dc3\u0db3\u0dc4\u0dcf \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2 \u0dc3\u0dc4 \u0db8\u0dd4\u0daf\u0dca\u0dbb\u0dd2\u0dad \u0db8\u0dad\u0d9a \u0dc3\u0dcf\u0dbb\u0dcf\u0d82\u0dc1\u0dba \u0db4\u0dd0\u0dc4\u0dd0\u0daf\u0dd2\u0dbd\u0dd2 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p><span translate=no>_^_0_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Device </p>\n": "<p>\u0d8b\u0db4\u0dcf\u0d82\u0d9c\u0dba </p>\n",
|
||||
"<p>Get next token. Note that we only feed the last token to the model because we cache the key/value pairs of previous tokens. </p>\n": "<p>\u0d8a\u0dc5\u0d9f\u0da7\u0ddd\u0d9a\u0db1\u0dba \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1. \u0db4\u0dd9\u0dbb \u0da7\u0ddd\u0d9a\u0db1 \u0dc0\u0dbd \u0dba\u0dad\u0dd4\u0dbb/\u0d85\u0d9c\u0dba \u0dba\u0dd4\u0d9c\u0dbd \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2 \u0d9a\u0dbb\u0db1 \u0db1\u0dd2\u0dc3\u0dcf \u0d85\u0db4\u0dd2 \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0da7 \u0d85\u0dc0\u0dc3\u0dcf\u0db1 \u0da7\u0ddd\u0d9a\u0db1\u0dba \u0db4\u0db8\u0dab\u0d9a\u0dca \u0db4\u0ddd\u0dc2\u0dab\u0dba \u0d9a\u0dbb\u0db1 \u0db6\u0dc0 \u0dc3\u0dbd\u0d9a\u0db1\u0dca\u0db1. </p>\n",
|
||||
"<p>Get token ids </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0dca\u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca\u0db4\u0dad\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Load layers in float16 into CPU. We convert the layers to int8 later, because doing that on the fly after loading layers to GPU causes CUDA memory fragmentation (about 3GB memory can get lost due to fragmentation). </p>\n": "<p>\u0db4\u0dcf\u0dc0\u0dd9\u0db116 \u0dc4\u0dd2 \u0dc3\u0dca\u0dae\u0dbb CPU \u0dad\u0dd4\u0dc5\u0da7 \u0db4\u0da7\u0dc0\u0db1\u0dca\u0db1. \u0d85\u0db4\u0dd2 \u0dc3\u0dca\u0dae\u0dbb \u0db4\u0dc3\u0dd4\u0dc0 int8 \u0db6\u0dc0\u0da7 \u0db4\u0dbb\u0dd2\u0dc0\u0dbb\u0dca\u0dad\u0db1\u0dba \u0d9a\u0dbb\u0db8\u0dd4, \u0db8\u0db1\u0dca\u0daf \u0dc3\u0dca\u0dae\u0dbb GPU \u0dc0\u0dd9\u0dad \u0db4\u0dd0\u0da7\u0dc0\u0dd3\u0db8\u0dd9\u0db1\u0dca \u0db4\u0dc3\u0dd4 \u0db4\u0dd2\u0dba\u0dcf\u0dc3\u0dbb \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 CUDA \u0db8\u0dad\u0d9a \u0d9b\u0dab\u0dca\u0da9\u0db1\u0dba \u0dc0\u0dd3\u0db8\u0da7 \u0dc4\u0dda\u0dad\u0dd4 \u0dc0\u0dda (3GB \u0db4\u0db8\u0dab \u0db8\u0dad\u0d9a\u0dba \u0d9a\u0dd0\u0db6\u0dbd\u0dd2 \u0dc0\u0dd3\u0db8 \u0db1\u0dd2\u0dc3\u0dcf \u0d85\u0dc4\u0dd2\u0db8\u0dd2 \u0dc0\u0dd2\u0dba \u0dc4\u0dd0\u0d9a). </p>\n",
|
||||
"<p>Predict 100 tokens </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1100 \u0d9a\u0dca \u0db4\u0dd4\u0dbb\u0ddd\u0d9a\u0dae\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Print </p>\n": "<p>\u0db8\u0dd4\u0daf\u0dca\u0dbb\u0dab\u0dba </p>\n",
|
||||
"<p>Run the model. We use the <a href=\"generate.html\"><span translate=no>_^_0_^_</span></a> function defined in <a href=\"generate.html\"><span translate=no>_^_1_^_</span></a> </p>\n": "<p>\u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0db0\u0dcf\u0dc0\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1. \u0d85\u0db4\u0dd2 \u0d85\u0dbb\u0dca\u0dae \u0daf\u0d9a\u0dca\u0dc0\u0dcf \u0d87\u0dad\u0dd2 <a href=\"generate.html\"><span translate=no>_^_0_^_</span></a> \u0dc1\u0dca\u0dbb\u0dd2\u0dad\u0dba \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd4 <a href=\"generate.html\"><span translate=no>_^_1_^_</span></a> </p>\n",
|
||||
"<p>Set the state to use cached activations </p>\n": "<p>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dc3\u0d9a\u0dca\u0dbb\u0dd2\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dca \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0da7 \u0dbb\u0dcf\u0da2\u0dca\u0dba\u0dba \u0dc3\u0d9a\u0dc3\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Setup <a href=\"../utils/cache.html\">cache</a> to cache intermediate key/value pairs for faster generation </p>\n": "<p>\u0dc0\u0dda\u0d9c\u0dc0\u0dad\u0dca\u0d8b\u0dad\u0dca\u0db4\u0dcf\u0daf\u0db1\u0dba \u0dc3\u0db3\u0dc4\u0dcf \u0d85\u0dad\u0dbb\u0db8\u0dd0\u0daf\u0dd2 \u0dba\u0dad\u0dd4\u0dbb/\u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0dca \u0dba\u0dd4\u0d9c\u0dbd <a href=\"../utils/cache.html\">\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2</a> \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0da7 \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba \u0db4\u0dd2\u0dc4\u0dd2\u0da7\u0dd4\u0dc0\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>This reduces CUDA memory fragmentation </p>\n": "<p>\u0db8\u0dd9\u0dbaCUDA \u0db8\u0dad\u0d9a \u0d9b\u0dab\u0dca\u0da9\u0db1\u0dba \u0d85\u0da9\u0dd4 \u0d9a\u0dbb\u0dba\u0dd2 </p>\n",
|
||||
"Generate Text with GPT-NeoX using LLM.int8() quantization": "LLM.INT8 () \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0d9a\u0dbb\u0dab\u0dba \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd2\u0db1\u0dca GPT-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dc3\u0db8\u0d9f \u0db4\u0dd9\u0dc5 \u0da2\u0db1\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h1>Generate Text with GPT-NeoX using LLM.int8() quantization</h1>\n<p>This shows how to generate text from GPT-NeoX using <a href=\"../utils/llm_int8.html\">LLM.int8() quantization</a>.</p>\n<p>This needs a GPU with 24GB memory.</p>\n": "<h1>\u4f7f\u7528 llm.int8 () \u91cf\u5316\u4f7f\u7528 GPT-NEOX \u751f\u6210\u6587\u672c</h1>\n<p>\u8fd9\u8bf4\u660e\u4e86\u5982\u4f55\u4f7f\u7528 <a href=\"../utils/llm_int8.html\">llm.int8 () \u91cf\u5316</a>\u4ece GPT-NEOX \u751f\u6210\u6587\u672c\u3002</p>\n<p>\u8fd9\u9700\u8981\u4e00\u4e2a\u5177\u6709 24GB \u5185\u5b58\u7684 GPU\u3002</p>\n",
|
||||
"<h2>Generate text</h2>\n": "<h2>\u751f\u6210\u6587\u672c</h2>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Append the predicted token </p>\n": "<p>\u8ffd\u52a0\u9884\u6d4b\u7684\u4ee4\u724c</p>\n",
|
||||
"<p>Clear cache and print memory summary for debugging </p>\n": "<p>\u6e05\u9664\u7f13\u5b58\u548c\u6253\u5370\u5185\u5b58\u6458\u8981\u4ee5\u8fdb\u884c\u8c03\u8bd5</p>\n",
|
||||
"<p>Create <span translate=no>_^_0_^_</span> model </p>\n": "<p>\u521b\u5efa<span translate=no>_^_0_^_</span>\u6a21\u578b</p>\n",
|
||||
"<p>Device </p>\n": "<p>\u8bbe\u5907</p>\n",
|
||||
"<p>Get next token. Note that we only feed the last token to the model because we cache the key/value pairs of previous tokens. </p>\n": "<p>\u83b7\u53d6\u4e0b\u4e00\u4e2a\u4ee4\u724c\u3002\u8bf7\u6ce8\u610f\uff0c\u6211\u4eec\u53ea\u5c06\u6700\u540e\u4e00\u4e2a\u4ee4\u724c\u63d0\u4f9b\u7ed9\u6a21\u578b\uff0c\u56e0\u4e3a\u6211\u4eec\u7f13\u5b58\u4e86\u5148\u524d\u4ee4\u724c\u7684\u952e/\u503c\u5bf9\u3002</p>\n",
|
||||
"<p>Get token ids </p>\n": "<p>\u83b7\u53d6\u4ee3\u5e01 ID</p>\n",
|
||||
"<p>Load layers in float16 into CPU. We convert the layers to int8 later, because doing that on the fly after loading layers to GPU causes CUDA memory fragmentation (about 3GB memory can get lost due to fragmentation). </p>\n": "\u5c06 <p>float16 \u4e2d\u7684\u5c42\u52a0\u8f7d\u5230 CPU \u4e2d\u3002\u6211\u4eec\u7a0d\u540e\u5c06\u56fe\u5c42\u8f6c\u6362\u4e3aint8\uff0c\u56e0\u4e3a\u5728\u5c06\u56fe\u5c42\u52a0\u8f7d\u5230GPU\u540e\u5373\u65f6\u6267\u884c\u6b64\u64cd\u4f5c\u4f1a\u5bfc\u81f4CUDA\u5185\u5b58\u788e\u7247\uff08\u5927\u7ea63GB\u7684\u5185\u5b58\u53ef\u80fd\u4f1a\u7531\u4e8e\u788e\u7247\u800c\u4e22\u5931\uff09\u3002</p>\n",
|
||||
"<p>Predict 100 tokens </p>\n": "<p>\u9884\u6d4b 100 \u4e2a\u4ee3\u5e01</p>\n",
|
||||
"<p>Print </p>\n": "<p>\u6253\u5370</p>\n",
|
||||
"<p>Run the model. We use the <a href=\"generate.html\"><span translate=no>_^_0_^_</span></a> function defined in <a href=\"generate.html\"><span translate=no>_^_1_^_</span></a> </p>\n": "<p>\u8fd0\u884c\u6a21\u578b\u3002\u6211\u4eec\u4f7f\u7528\u4e2d\u5b9a\u4e49\u7684<a href=\"generate.html\"><span translate=no>_^_0_^_</span></a>\u51fd\u6570 <a href=\"generate.html\"><span translate=no>_^_1_^_</span></a></p>\n",
|
||||
"<p>Set the state to use cached activations </p>\n": "<p>\u8bbe\u7f6e\u72b6\u6001\u4ee5\u4f7f\u7528\u7f13\u5b58\u7684\u6fc0\u6d3b</p>\n",
|
||||
"<p>Setup <a href=\"../utils/cache.html\">cache</a> to cache intermediate key/value pairs for faster generation </p>\n": "<p>\u8bbe\u7f6e<a href=\"../utils/cache.html\">\u7f13\u5b58</a>\u4ee5\u7f13\u5b58\u4e2d\u95f4\u952e/\u503c\u5bf9\u4ee5\u52a0\u5feb\u751f\u6210\u901f\u5ea6</p>\n",
|
||||
"<p>This reduces CUDA memory fragmentation </p>\n": "<p>\u8fd9\u51cf\u5c11\u4e86 CUDA \u5185\u5b58\u788e\u7247</p>\n",
|
||||
"Generate Text with GPT-NeoX using LLM.int8() quantization": "\u4f7f\u7528 llm.int8 () \u91cf\u5316\u4f7f\u7528 GPT-NEOX \u751f\u6210\u6587\u672c"
|
||||
}
|
||||
@@ -0,0 +1,6 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Tokenizer</h1>\n<p>This initializes a Hugging Face tokenizer from the downloaded vocabulary.</p>\n": "<h1>GPT-\u30cd\u30aa\u30c3\u30af\u30b9\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc</h1>\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3057\u305f\u30dc\u30ad\u30e3\u30d6\u30e9\u30ea\u30fc\u304b\u3089 Hugging Face \u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u304c\u521d\u671f\u5316\u3055\u308c\u307e\u3059\u3002</p>\n",
|
||||
"<h3>Load NeoX Tokenizer</h3>\n<ul><p><em>Returns</em> the tokenizer</p></ul>\n": "<h3>NeoX \u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u30ed\u30fc\u30c9</h3>\n<ul><p><em>\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"GPT-NeoX Tokenizer": "GPT-\u30cd\u30aa\u30c3\u30af\u30b9\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc",
|
||||
"Loads the GPT-NeoX tokenizer": "GPT-Neox \u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u30ed\u30fc\u30c9\u3057\u307e\u3059"
|
||||
}
|
||||
@@ -0,0 +1,6 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Tokenizer</h1>\n<p>This initializes a Hugging Face tokenizer from the downloaded vocabulary.</p>\n": "<h1>\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca</h1>\n<p>\u0db8\u0dd9\u0dba\u0db6\u0dcf\u0d9c\u0dad \u0d9a\u0dc5 \u0dc0\u0da0\u0db1 \u0db8\u0dcf\u0dbd\u0dcf\u0dc0\u0dd9\u0db1\u0dca \u0dc4\u0dd4\u0da2\u0dd2\u0d82 \u0dc6\u0dda\u0dc3\u0dca \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca \u0d86\u0dbb\u0db8\u0dca\u0db7 \u0d9a\u0dbb\u0dba\u0dd2. </p>\n",
|
||||
"<h3>Load NeoX Tokenizer</h3>\n<ul><p><em>Returns</em> the tokenizer</p></ul>\n": "<h3>Neox\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca \u0db4\u0dd0\u0da7\u0dc0\u0dd3\u0db8</h3>\n<ul><p>\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca<em>\u0d86\u0db4\u0dc3\u0dd4 \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"GPT-NeoX Tokenizer": "\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca",
|
||||
"Loads the GPT-NeoX tokenizer": "\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca \u0db4\u0da7\u0dc0\u0db1\u0dd4 \u0dbd\u0dd0\u0db6\u0dda"
|
||||
}
|
||||
@@ -0,0 +1,6 @@
|
||||
{
|
||||
"<h1>GPT-NeoX Tokenizer</h1>\n<p>This initializes a Hugging Face tokenizer from the downloaded vocabulary.</p>\n": "<h1>GPT-neox Tokenizer</h1>\n<p>\u8fd9\u4f1a\u4ece\u4e0b\u8f7d\u7684\u8bcd\u6c47\u8868\u4e2d\u521d\u59cb\u5316\u4e00\u4e2a Hugging Face \u5206\u8bcd\u5668\u3002</p>\n",
|
||||
"<h3>Load NeoX Tokenizer</h3>\n<ul><p><em>Returns</em> the tokenizer</p></ul>\n": "<h3>\u52a0\u8f7d neOX \u5206\u8bcd\u5668</h3>\n<ul><p><em>\u8fd4\u56de</em>\u5206\u8bcd\u5668</p></ul>\n",
|
||||
"GPT-NeoX Tokenizer": "GPT-neox Tokenizer",
|
||||
"Loads the GPT-NeoX tokenizer": "\u52a0\u8f7d GPT-NEOX \u5206\u8bcd\u5668"
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
{
|
||||
"<h1>Utilities and Helpers</h1>\n<ul><li><a href=\"cache.html\">Cache for intermediate activations (for faster inference)</a> </li>\n<li><a href=\"finetune.html\">Tools for finetuning</a> </li>\n<li><a href=\"trainer.html\">Trainer</a> </li>\n<li><a href=\"text_dataset.html\">Text dataset</a></li></ul>\n": "<h1>\u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3\u3068\u30d8\u30eb\u30d1\u30fc</h1>\n<ul><li><a href=\"cache.html\">\u4e2d\u9593\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u7528\u306e\u30ad\u30e3\u30c3\u30b7\u30e5 (\u63a8\u8ad6\u3092\u9ad8\u901f\u5316\u3059\u308b\u305f\u3081)</a></li>\n<li><a href=\"finetune.html\">\u5fae\u8abf\u6574\u7528\u30c4\u30fc\u30eb</a></li>\n<li><a href=\"trainer.html\">\u30c8\u30ec\u30fc\u30ca\u30fc</a></li>\n<li><a href=\"text_dataset.html\">\u30c6\u30ad\u30b9\u30c8\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8</a></li></ul>\n",
|
||||
"<h3>Balance layers</h3>\n<p>Split the <span translate=no>_^_0_^_</span> into <span translate=no>_^_1_^_</span>. This is used for pipeline parallel training.</p>\n<ul><li><span translate=no>_^_2_^_</span> is the number of layers </li>\n<li><span translate=no>_^_3_^_</span> is the number of chunks </li>\n<p><em>Returns</em> returns a list with the number of layers for each chunk</p></ul>\n": "<h3>\u30d0\u30e9\u30f3\u30b9\u30ec\u30a4\u30e4\u30fc</h3>\n<p>\u306b\u5206\u5272<span translate=no>_^_1_^_</span>. <span translate=no>_^_0_^_</span>\u3053\u308c\u306f\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u306e\u4e26\u5217\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u306b\u4f7f\u7528\u3055\u308c\u307e\u3059\u3002</p>\n<ul><li><span translate=no>_^_2_^_</span>\u306f\u30ec\u30a4\u30e4\u30fc\u306e\u6570</li>\n<li><span translate=no>_^_3_^_</span>\u306f\u30c1\u30e3\u30f3\u30af\u306e\u6570</li>\n<p>Returns \u306f\u3001<em>\u5404\u30c1\u30e3\u30f3\u30af\u306e\u30ec\u30a4\u30e4\u30fc\u6570\u306e\u30ea\u30b9\u30c8\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h3>Get token ids</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the text to tokenize </li>\n<p><em>Returns</em> the token ids</p></ul>\n": "<h3>\u30c8\u30fc\u30af\u30f3 ID \u3092\u53d6\u5f97</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30c8\u30fc\u30af\u30f3\u5316\u3059\u308b\u30c6\u30ad\u30b9\u30c8\u3067\u3059</li>\n<p>\u30c8\u30fc\u30af\u30f3 ID <em>\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h3>Print tokens from model outputs</h3>\n<p>Pretty prints target tokens along side outputs from the model(s).</p>\n<ul><li><span translate=no>_^_0_^_</span> are the target token ids </li>\n<li><span translate=no>_^_1_^_</span> are the model(s) outputs</li></ul>\n": "<h3>\u30e2\u30c7\u30eb\u51fa\u529b\u304b\u3089\u30c8\u30fc\u30af\u30f3\u3092\u5370\u5237</h3>\n<p>Pretty \u306f\u3001\u30e2\u30c7\u30eb\u304b\u3089\u306e\u51fa\u529b\u3068\u4e00\u7dd2\u306b\u30bf\u30fc\u30b2\u30c3\u30c8\u30c8\u30fc\u30af\u30f3\u3092\u51fa\u529b\u3057\u307e\u3059\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u30bf\u30fc\u30b2\u30c3\u30c8\u30c8\u30fc\u30af\u30f3 ID</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30e2\u30c7\u30eb\u306e\u51fa\u529b\u3067\u3059</li></ul>\n",
|
||||
"<h3>Print tokens</h3>\n<p>Pretty prints tokens for comparison</p>\n<ul><li><span translate=no>_^_0_^_</span> are the target token ids </li>\n<li><span translate=no>_^_1_^_</span> are the sampled outputs from the model(s)</li></ul>\n": "<h3>\u30c8\u30fc\u30af\u30f3\u306e\u5370\u5237</h3>\n<p>Pretty \u306f\u6bd4\u8f03\u7528\u306e\u30c8\u30fc\u30af\u30f3\u3092\u5370\u5237\u3057\u307e\u3059</p>\n<ul><li><span translate=no>_^_0_^_</span>\u30bf\u30fc\u30b2\u30c3\u30c8\u30c8\u30fc\u30af\u30f3 ID</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30e2\u30c7\u30eb\u304b\u3089\u30b5\u30f3\u30d7\u30ea\u30f3\u30b0\u3055\u308c\u305f\u51fa\u529b\u3067\u3059</li></ul>\n",
|
||||
"<p>Convert the tokens to list of strings </p>\n": "<p>\u30c8\u30fc\u30af\u30f3\u3092\u6587\u5b57\u5217\u306e\u30ea\u30b9\u30c8\u306b\u5909\u63db</p>\n",
|
||||
"<p>Empty target </p>\n": "<p>\u7a7a\u306e\u30bf\u30fc\u30b2\u30c3\u30c8</p>\n",
|
||||
"<p>Iterate through tokens </p>\n": "<p>\u30c8\u30fc\u30af\u30f3\u3092\u7e70\u308a\u8fd4\u3057\u51e6\u7406</p>\n",
|
||||
"<p>Load tokenizer </p>\n": "<p>\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u30ed\u30fc\u30c9</p>\n",
|
||||
"<p>Number of tokens </p>\n": "<p>\u30c8\u30fc\u30af\u30f3\u306e\u6570</p>\n",
|
||||
"<p>Other outputs </p>\n": "<p>\u305d\u306e\u4ed6\u306e\u51fa\u529b</p>\n",
|
||||
"<p>Stats </p>\n": "<p>\u7d71\u8a08\u60c5\u5831</p>\n",
|
||||
"<p>Tokenizer singleton </p>\n": "<p>\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u30b7\u30f3\u30b0\u30eb\u30c8\u30f3</p>\n",
|
||||
"Utilities and Helpers": "\u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3\u3068\u30d8\u30eb\u30d1\u30fc",
|
||||
"Utilities and helper functions": "\u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3\u3068\u30d8\u30eb\u30d1\u30fc\u95a2\u6570"
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
{
|
||||
"<h1>Utilities and Helpers</h1>\n<ul><li><a href=\"cache.html\">Cache for intermediate activations (for faster inference)</a> </li>\n<li><a href=\"finetune.html\">Tools for finetuning</a> </li>\n<li><a href=\"trainer.html\">Trainer</a> </li>\n<li><a href=\"text_dataset.html\">Text dataset</a></li></ul>\n": "<h1>\u0d8b\u0db4\u0dba\u0ddd\u0d9c\u0dd2\u0dad\u0dcf\u0dc3\u0dc4 \u0d8b\u0daf\u0dc0\u0dca\u0d9a\u0dbb\u0dd4\u0dc0\u0db1\u0dca</h1>\n<ul><li><a href=\"cache.html\">\u0d85\u0dad\u0dbb\u0db8\u0dd0\u0daf\u0dd2 \u0dc3\u0d9a\u0dca\u0dbb\u0dd2\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dca \u0dc3\u0db3\u0dc4\u0dcf \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba (\u0dc0\u0dda\u0d9c\u0dc0\u0dad\u0dca \u0d85\u0db1\u0dd4\u0db8\u0dcf\u0db1\u0dba \u0dc3\u0db3\u0dc4\u0dcf)</a> </li>\n<li><a href=\"finetune.html\">\u0d85\u0dc0\u0dc3\u0db1\u0dca \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0db8\u0dd9\u0dc0\u0dbd\u0db8\u0dca</a> </li>\n<li><a href=\"trainer.html\">\u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4\u0d9a\u0dbb\u0dd4</a> </li>\n<li><a href=\"text_dataset.html\">\u0db4\u0dd9\u0dc5 \u0daf\u0dad\u0dca\u0dad \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba</a></li></ul>\n",
|
||||
"<h3>Balance layers</h3>\n<p>Split the <span translate=no>_^_0_^_</span> into <span translate=no>_^_1_^_</span>. This is used for pipeline parallel training.</p>\n<ul><li><span translate=no>_^_2_^_</span> is the number of layers </li>\n<li><span translate=no>_^_3_^_</span> is the number of chunks </li>\n<p><em>Returns</em> returns a list with the number of layers for each chunk</p></ul>\n": "<h3>\u0dc1\u0dda\u0dc2\u0dc3\u0dca\u0dae\u0dbb</h3>\n<p>\u0db6\u0dd9\u0daf\u0db1\u0dca\u0db1 <span translate=no>_^_1_^_</span>. <span translate=no>_^_0_^_</span> \u0db1\u0dbd \u0db8\u0dcf\u0dbb\u0dca\u0d9c \u0dc3\u0db8\u0dcf\u0db1\u0dca\u0dad\u0dbb \u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4\u0dc0 \u0dc3\u0db3\u0dc4\u0dcf \u0db8\u0dd9\u0dba \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0dc0\u0dda. </p>\n<ul><li><span translate=no>_^_2_^_</span> \u0dc3\u0dca\u0dae\u0dbb \u0d9c\u0dab\u0db1 \u0dc0\u0dda </li>\n<li><span translate=no>_^_3_^_</span> \u0d9a\u0dd4\u0da7\u0dca\u0da7\u0dd2 \u0d9c\u0dab\u0db1 \u0dc0\u0dda </li>\n</ul><p><em>\u0db4\u0dca\u0dbb\u0dad\u0dd2\u0dbd\u0dcf\u0db7</em> \u0d91\u0d9a\u0dca \u0d91\u0d9a\u0dca \u0d9a\u0dd4\u0da7\u0dca\u0da7\u0dd2\u0dba \u0dc3\u0db3\u0dc4\u0dcf \u0dc3\u0dca\u0dae\u0dbb \u0d9c\u0dab\u0db1 \u0dc3\u0dc4\u0dd2\u0dad \u0dbd\u0dd0\u0dba\u0dd2\u0dc3\u0dca\u0dad\u0dd4\u0dc0\u0d9a\u0dca \u0d86\u0db4\u0dc3\u0dd4 \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</p>\n",
|
||||
"<h3>Get token ids</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the text to tokenize </li>\n<p><em>Returns</em> the token ids</p></ul>\n": "<h3>\u0da7\u0ddd\u0d9a\u0db1\u0dca\u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca\u0db4\u0dad\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dca \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda \u0db4\u0dcf text \u0dba \u0dc0\u0dda </li>\n<p>\u0da7\u0ddd\u0d9a\u0db1\u0dca\u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca\u0db4\u0dad\u0dca<em>\u0d86\u0db4\u0dc3\u0dd4 \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<h3>Print tokens from model outputs</h3>\n<p>Pretty prints target tokens along side outputs from the model(s).</p>\n<ul><li><span translate=no>_^_0_^_</span> are the target token ids </li>\n<li><span translate=no>_^_1_^_</span> are the model(s) outputs</li></ul>\n": "<h3>\u0d86\u0daf\u0dbb\u0dca\u0dc1\u0db1\u0dd2\u0db8\u0dd0\u0dc0\u0dd4\u0db8\u0dca \u0dc0\u0dbd\u0dd2\u0db1\u0dca \u0da7\u0ddd\u0d9a\u0db1 \u0db8\u0dd4\u0daf\u0dca\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<p>\u0dbd\u0dc3\u0dca\u0dc3\u0db1\u0db8\u0dd4\u0daf\u0dca\u0dbb\u0dab \u0db8\u0d9f\u0dd2\u0db1\u0dca \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0dda (\u0dba) \u0dc3\u0dd2\u0da7 \u0db4\u0dd0\u0dad\u0dd2 \u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba\u0db1\u0dca \u0d94\u0dc3\u0dca\u0dc3\u0dda \u0da7\u0ddd\u0d9a\u0db1 \u0d89\u0dbd\u0d9a\u0dca\u0d9a \u0d9a\u0dbb\u0dba\u0dd2. </p>\n<ul><li><span translate=no>_^_0_^_</span> \u0d89\u0dbd\u0d9a\u0dca\u0d9a\u0d9c\u0dad \u0da7\u0ddd\u0d9a\u0db1\u0dca \u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca \u0dc0\u0dda </li>\n</ul><li><span translate=no>_^_1_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba (\u0dba) \u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba\u0db1\u0dca \u0dc0\u0dda</li>\n",
|
||||
"<h3>Print tokens</h3>\n<p>Pretty prints tokens for comparison</p>\n<ul><li><span translate=no>_^_0_^_</span> are the target token ids </li>\n<li><span translate=no>_^_1_^_</span> are the sampled outputs from the model(s)</li></ul>\n": "<h3>\u0da7\u0ddd\u0d9a\u0db1\u0db8\u0dd4\u0daf\u0dca\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<p>\u0dc3\u0d82\u0dc3\u0db1\u0dca\u0daf\u0db1\u0dba\u0dc3\u0db3\u0dc4\u0dcf \u0da7\u0ddd\u0d9a\u0db1 \u0db8\u0dd4\u0daf\u0dca\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</p>\n<ul><li><span translate=no>_^_0_^_</span> \u0d89\u0dbd\u0d9a\u0dca\u0d9a\u0d9c\u0dad \u0da7\u0ddd\u0d9a\u0db1\u0dca \u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0dd9\u0db1\u0dca \u0db1\u0dd2\u0dba\u0dd0\u0daf\u0dd2 \u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba\u0db1\u0dca (\u0dba)</li></ul>\n",
|
||||
"<p>Convert the tokens to list of strings </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0db1\u0dd6\u0dbd\u0dca \u0dbd\u0dd0\u0dba\u0dd2\u0dc3\u0dca\u0dad\u0dd4\u0dc0\u0da7 \u0db4\u0dbb\u0dd2\u0dc0\u0dbb\u0dca\u0dad\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Empty target </p>\n": "<p>\u0dc4\u0dd2\u0dc3\u0dca\u0d89\u0dbd\u0d9a\u0dca\u0d9a\u0dba </p>\n",
|
||||
"<p>Iterate through tokens </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0dc4\u0dbb\u0dc4\u0dcf \u0db1\u0dd0\u0dc0\u0dad \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Load tokenizer </p>\n": "<p>\u0db6\u0dbb\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca </p>\n",
|
||||
"<p>Number of tokens </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0d9c\u0dab\u0db1 </p>\n",
|
||||
"<p>Other outputs </p>\n": "<p>\u0dc0\u0dd9\u0db1\u0dad\u0dca\u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba\u0db1\u0dca </p>\n",
|
||||
"<p>Stats </p>\n": "<p>\u0dc3\u0d82\u0d9b\u0dca\u0dba\u0dcf\u0db1 </p>\n",
|
||||
"<p>Tokenizer singleton </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dbb\u0dca\u0dc3\u0dd2\u0d82\u0d9c\u0dbd\u0dca\u0da7\u0db1\u0dca </p>\n",
|
||||
"Utilities and Helpers": "\u0d8b\u0db4\u0dba\u0ddd\u0d9c\u0dd2\u0dad\u0dcf \u0dc3\u0dc4 \u0d8b\u0daf\u0dc0\u0dca\u0d9a\u0dbb\u0dd4\u0dc0\u0db1\u0dca",
|
||||
"Utilities and helper functions": "\u0d8b\u0db4\u0dba\u0ddd\u0d9c\u0dd2\u0dad\u0dcf \u0dc3\u0dc4 \u0d8b\u0db4\u0d9a\u0dcf\u0dbb\u0d9a \u0d9a\u0dcf\u0dbb\u0dca\u0dba\u0dba\u0db1\u0dca"
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
{
|
||||
"<h1>Utilities and Helpers</h1>\n<ul><li><a href=\"cache.html\">Cache for intermediate activations (for faster inference)</a> </li>\n<li><a href=\"finetune.html\">Tools for finetuning</a> </li>\n<li><a href=\"trainer.html\">Trainer</a> </li>\n<li><a href=\"text_dataset.html\">Text dataset</a></li></ul>\n": "<h1>\u516c\u7528\u4e8b\u4e1a\u548c\u52a9\u624b</h1>\n<ul><li><a href=\"cache.html\">\u7f13\u5b58\u7528\u4e8e\u4e2d\u95f4\u6fc0\u6d3b\uff08\u7528\u4e8e\u66f4\u5feb\u7684\u63a8\u7406\uff09</a></li>\n<li><a href=\"finetune.html\">\u5fae\u8c03\u5de5\u5177</a></li>\n<li><a href=\"trainer.html\">\u8bad\u7ec3\u5e08</a></li>\n<li><a href=\"text_dataset.html\">\u6587\u672c\u6570\u636e\u96c6</a></li></ul>\n",
|
||||
"<h3>Balance layers</h3>\n<p>Split the <span translate=no>_^_0_^_</span> into <span translate=no>_^_1_^_</span>. This is used for pipeline parallel training.</p>\n<ul><li><span translate=no>_^_2_^_</span> is the number of layers </li>\n<li><span translate=no>_^_3_^_</span> is the number of chunks </li>\n<p><em>Returns</em> returns a list with the number of layers for each chunk</p></ul>\n": "<h3>\u5e73\u8861\u56fe\u5c42</h3>\n<p>\u62c6\u5206<span translate=no>_^_0_^_</span>\u6210<span translate=no>_^_1_^_</span>\u3002\u8fd9\u7528\u4e8e\u7ba1\u9053\u5e76\u884c\u8bad\u7ec3\u3002</p>\n<ul><li><span translate=no>_^_2_^_</span>\u662f\u5c42\u6570</li>\n<li><span translate=no>_^_3_^_</span>\u662f\u533a\u5757\u7684\u6570\u91cf</li>\n<p><em>\u8fd4\u56de</em>\u4e00\u4e2a\u5217\u8868\uff0c\u5176\u4e2d\u5305\u542b\u6bcf\u4e2a\u533a\u5757\u7684\u5c42\u6570</p></ul>\n",
|
||||
"<h3>Get token ids</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the text to tokenize </li>\n<p><em>Returns</em> the token ids</p></ul>\n": "<h3>\u83b7\u53d6\u4ee3\u5e01 ID</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u8981\u6807\u8bb0\u7684\u6587\u672c</li>\n<p><em>\u8fd4\u56de</em>\u4ee4\u724c ID</p></ul>\n",
|
||||
"<h3>Print tokens from model outputs</h3>\n<p>Pretty prints target tokens along side outputs from the model(s).</p>\n<ul><li><span translate=no>_^_0_^_</span> are the target token ids </li>\n<li><span translate=no>_^_1_^_</span> are the model(s) outputs</li></ul>\n": "<h3>\u4ece\u6a21\u578b\u8f93\u51fa\u4e2d\u6253\u5370\u4ee4\u724c</h3>\nP@@ <p>retty \u6cbf\u7740\u6a21\u578b\u7684\u4fa7\u9762\u8f93\u51fa\u6253\u5370\u76ee\u6807\u4ee4\u724c\u3002</p>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u76ee\u6807\u4ee3\u5e01 ID</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u6a21\u578b\u7684\u8f93\u51fa</li></ul>\n",
|
||||
"<h3>Print tokens</h3>\n<p>Pretty prints tokens for comparison</p>\n<ul><li><span translate=no>_^_0_^_</span> are the target token ids </li>\n<li><span translate=no>_^_1_^_</span> are the sampled outputs from the model(s)</li></ul>\n": "<h3>\u6253\u5370\u4ee3\u5e01</h3>\n<p>Pretty \u6253\u5370\u4ee4\u724c\u8fdb\u884c\u6bd4\u8f83</p>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u76ee\u6807\u4ee3\u5e01 ID</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u6765\u81ea\u6a21\u578b\u7684\u91c7\u6837\u8f93\u51fa</li></ul>\n",
|
||||
"<p>Convert the tokens to list of strings </p>\n": "<p>\u5c06\u6807\u8bb0\u8f6c\u6362\u4e3a\u5b57\u7b26\u4e32\u5217\u8868</p>\n",
|
||||
"<p>Empty target </p>\n": "<p>\u7a7a\u76ee\u6807</p>\n",
|
||||
"<p>Iterate through tokens </p>\n": "<p>\u904d\u5386\u4ee4\u724c</p>\n",
|
||||
"<p>Load tokenizer </p>\n": "<p>\u52a0\u8f7d\u5206\u8bcd\u5668</p>\n",
|
||||
"<p>Number of tokens </p>\n": "<p>\u4ee3\u5e01\u6570\u91cf</p>\n",
|
||||
"<p>Other outputs </p>\n": "<p>\u5176\u4ed6\u8f93\u51fa</p>\n",
|
||||
"<p>Stats </p>\n": "<p>\u7edf\u8ba1\u6570\u636e</p>\n",
|
||||
"<p>Tokenizer singleton </p>\n": "<p>\u5206\u8bcd\u5668\u5355\u4f8b</p>\n",
|
||||
"Utilities and Helpers": "\u516c\u7528\u4e8b\u4e1a\u548c\u52a9\u624b",
|
||||
"Utilities and helper functions": "\u5b9e\u7528\u7a0b\u5e8f\u548c\u8f85\u52a9\u51fd\u6570"
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
{
|
||||
"<h1>Cache for Intermediate Activations</h1>\n<p>During inference the model outputs token by token. We use this simple cache to store key's and value's attention layers, so that we don't have to recompute them for previous tokens.</p>\n": "<h1>\u4e2d\u9593\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u7528\u30ad\u30e3\u30c3\u30b7\u30e5</h1>\n<p>\u63a8\u8ad6\u4e2d\u3001\u30e2\u30c7\u30eb\u306f\u30c8\u30fc\u30af\u30f3\u3054\u3068\u306b\u51fa\u529b\u3057\u307e\u3059\u3002\u3053\u306e\u30b7\u30f3\u30d7\u30eb\u306a\u30ad\u30e3\u30c3\u30b7\u30e5\u3092\u4f7f\u3063\u3066\u30ad\u30fc\u3068\u5024\u306e\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30ec\u30a4\u30e4\u30fc\u3092\u683c\u7d0d\u3059\u308b\u306e\u3067\u3001\u4ee5\u524d\u306e\u30c8\u30fc\u30af\u30f3\u3067\u305d\u308c\u3089\u3092\u518d\u8a08\u7b97\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u305b\u3093</p>\u3002\n",
|
||||
"<h2>Cache</h2>\n<p>This maintains a key-value cache and queues push values and pop them in the same order. The queues are useful since we have multiple attention layers.</p>\n": "<h2>\u30ad\u30e3\u30c3\u30b7\u30e5</h2>\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30ad\u30fc\u3068\u5024\u306e\u30ad\u30e3\u30c3\u30b7\u30e5\u304c\u7dad\u6301\u3055\u308c\u3001\u5024\u306e\u30d7\u30c3\u30b7\u30e5\u3068\u30dd\u30c3\u30d7\u304c\u540c\u3058\u9806\u5e8f\u3067\u30ad\u30e5\u30fc\u306b\u5165\u308c\u3089\u308c\u307e\u3059\u3002\u30ad\u30e5\u30fc\u306f\u8907\u6570\u306e\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30ec\u30a4\u30e4\u30fc\u304c\u3042\u308b\u306e\u3067\u4fbf\u5229\u3067\u3059</p>\u3002\n",
|
||||
"<h3>Cache a value</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the value to be cached </li>\n<li><span translate=no>_^_1_^_</span> is the value</li></ul>\n": "<h3>\u5024\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3059\u308b</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30ad\u30e3\u30c3\u30b7\u30e5\u3055\u308c\u308b\u5024\u306e\u540d\u524d</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u5024\u3067\u3059</li></ul>\n",
|
||||
"<h3>Clear a cache value</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name used when caching</li></ul>\n": "<h3>\u30ad\u30e3\u30c3\u30b7\u30e5\u5024\u3092\u30af\u30ea\u30a2\u3059\u308b</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30ad\u30e3\u30c3\u30b7\u30e5\u6642\u306b\u4f7f\u7528\u3055\u308c\u308b\u540d\u524d\u3067\u3059</li></ul>\n",
|
||||
"<h3>Clear cache</h3>\n": "<h3>\u30ad\u30e3\u30c3\u30b7\u30e5\u3092\u30af\u30ea\u30a2</h3>\n",
|
||||
"<h3>Get the cache instance</h3>\n<ul><p><em>Returns</em> the cache instance</p></ul>\n": "<h3>\u30ad\u30e3\u30c3\u30b7\u30e5\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3092\u53d6\u5f97</h3>\n<ul><p><em>\u30ad\u30e3\u30c3\u30b7\u30e5\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h3>Pop from a queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<p><em>Returns</em> the value</p></ul>\n": "<h3>\u30ad\u30e5\u30fc\u304b\u3089\u30dd\u30c3\u30d7\u3059\u308b</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u306f\u30ad\u30e5\u30fc\u306e\u540d\u524d\u3067\u3059</li>\n<p><em>\u5024\u3092\u8fd4\u3059</em></p></ul>\n",
|
||||
"<h3>Push a value to a queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<li><span translate=no>_^_1_^_</span> is the value to be pushed</li></ul>\n": "<h3>\u5024\u3092\u30ad\u30e5\u30fc\u306b\u30d7\u30c3\u30b7\u30e5</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u306f\u30ad\u30e5\u30fc\u306e\u540d\u524d\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u30d7\u30c3\u30b7\u30e5\u3059\u308b\u5024\u3067\u3059</li></ul>\n",
|
||||
"<h3>Retrieve a value from cache</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name used when caching </li>\n<li><span translate=no>_^_1_^_</span> is the default value if the cache is empty </li>\n<p><em>Returns</em> the cached value</p></ul>\n": "<h3>\u30ad\u30e3\u30c3\u30b7\u30e5\u304b\u3089\u5024\u3092\u53d6\u5f97</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30ad\u30e3\u30c3\u30b7\u30e5\u6642\u306b\u4f7f\u7528\u3055\u308c\u308b\u540d\u524d\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u30ad\u30e3\u30c3\u30b7\u30e5\u304c\u7a7a\u306e\u5834\u5408\u306e\u30c7\u30d5\u30a9\u30eb\u30c8\u5024\u3067\u3059</li>\n<p><em>\u30ad\u30e3\u30c3\u30b7\u30e5\u3055\u308c\u305f\u5024\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h3>Return the size of the queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<p><em>Returns</em> size of the queue if exists else None</p></ul>\n": "<h3>\u30ad\u30e5\u30fc\u306e\u30b5\u30a4\u30ba\u3092\u8fd4\u3059</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u306f\u30ad\u30e5\u30fc\u306e\u540d\u524d\u3067\u3059</li>\n<p><em>\u30ad\u30e5\u30fc\u304c\u5b58\u5728\u3059\u308b\u5834\u5408\u306f\u30ad\u30e5\u30fc\u306e\u30b5\u30a4\u30ba\u3092\u8fd4\u3057\u307e\u3059</em>\u3002\u305d\u308c\u4ee5\u5916\u306e\u5834\u5408\u306f None</p></ul>\n",
|
||||
"<p>Create an empty queue if it's not present </p>\n": "<p>\u5b58\u5728\u3057\u306a\u3044\u5834\u5408\u306f\u7a7a\u306e\u30ad\u30e5\u30fc\u3092\u4f5c\u6210</p>\n",
|
||||
"<p>Push to the queue </p>\n": "<p>\u30ad\u30e5\u30fc\u306b\u30d7\u30c3\u30b7\u30e5</p>\n",
|
||||
"<p>Singleton for cache </p>\n": "<p>\u30ad\u30e3\u30c3\u30b7\u30e5\u7528\u30b7\u30f3\u30b0\u30eb\u30c8\u30f3</p>\n",
|
||||
"Cache for Intermediate Activations": "\u4e2d\u9593\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u7528\u30ad\u30e3\u30c3\u30b7\u30e5",
|
||||
"Cache for intermediate activations for faster inference.": "\u63a8\u8ad6\u3092\u9ad8\u901f\u5316\u3059\u308b\u305f\u3081\u306e\u4e2d\u9593\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u7528\u306e\u30ad\u30e3\u30c3\u30b7\u30e5\u3002"
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
{
|
||||
"<h1>Cache for Intermediate Activations</h1>\n<p>During inference the model outputs token by token. We use this simple cache to store key's and value's attention layers, so that we don't have to recompute them for previous tokens.</p>\n": "<h1>\u0d85\u0dad\u0dbb\u0db8\u0dd0\u0daf\u0dd2\u0dc3\u0d9a\u0dca\u0dbb\u0dd3\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dca \u0dc3\u0db3\u0dc4\u0dcf \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba</h1>\n<p>\u0d85\u0db1\u0dd4\u0db8\u0dcf\u0db1\u0dba\u0d85\u0dad\u0dbb\u0dad\u0dd4\u0dbb \u0d86\u0daf\u0dbb\u0dca\u0dc1 \u0da7\u0ddd\u0d9a\u0db1\u0dba \u0db8\u0d9f\u0dd2\u0db1\u0dca \u0da7\u0ddd\u0d9a\u0db1\u0dba \u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba \u0d9a\u0dbb\u0dba\u0dd2. \u0dba\u0dad\u0dd4\u0dbb\u0dda \u0dc3\u0dc4 \u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0dda \u0d85\u0dc0\u0db0\u0dcf\u0db1\u0dba \u0dc3\u0dca\u0dae\u0dbb \u0d9c\u0db6\u0da9\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0d85\u0db4\u0dd2 \u0db8\u0dd9\u0db8 \u0dc3\u0dbb\u0dbd \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db8\u0dd4, \u0d91\u0dc0\u0dd2\u0da7 \u0db4\u0dd9\u0dbb \u0da7\u0ddd\u0d9a\u0db1 \u0dc3\u0db3\u0dc4\u0dcf \u0d92\u0dc0\u0dcf \u0db1\u0dd0\u0dc0\u0dad \u0d9c\u0dab\u0db1\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0da7 \u0d85\u0db4\u0da7 \u0d85\u0dc0\u0dc1\u0dca\u0dba \u0db1\u0ddc\u0dc0\u0dda. </p>\n",
|
||||
"<h2>Cache</h2>\n<p>This maintains a key-value cache and queues push values and pop them in the same order. The queues are useful since we have multiple attention layers.</p>\n": "<h2>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba</h2>\n<p>\u0db8\u0dd9\u0dba\u0db4\u0dca\u0dbb\u0db0\u0dcf\u0db1 \u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0d9a\u0dca \u0dc3\u0dc4\u0dd2\u0dad \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba\u0d9a\u0dca \u0db4\u0dc0\u0dad\u0dca\u0dc0\u0dcf \u0d9c\u0dd9\u0db1 \u0dba\u0db1 \u0d85\u0dad\u0dbb \u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dca \u0d85\u0d9c\u0dba\u0db1\u0dca \u0dad\u0dbd\u0dca\u0dbd\u0dd4 \u0d9a\u0dbb \u0d92\u0dc0\u0dcf \u0d91\u0d9a\u0db8 \u0d85\u0db1\u0dd4\u0db4\u0dd2\u0dc5\u0dd2\u0dc0\u0dd9\u0dbd\u0d9a\u0da7 \u0db4\u0ddc\u0db4\u0dca \u0d9a\u0dbb\u0dba\u0dd2. \u0d85\u0db4\u0da7 \u0db6\u0dc4\u0dd4 \u0d85\u0dc0\u0db0\u0dcf\u0db1\u0dba \u0dc3\u0dca\u0dae\u0dbb \u0d87\u0dad\u0dd2 \u0db6\u0dd0\u0dc0\u0dd2\u0db1\u0dca \u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dca \u0db4\u0dca\u0dbb\u0dba\u0ddd\u0da2\u0db1\u0dc0\u0dad\u0dca \u0dc0\u0dda. </p>\n",
|
||||
"<h3>Cache a value</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the value to be cached </li>\n<li><span translate=no>_^_1_^_</span> is the value</li></ul>\n": "<h3>\u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0d9a\u0dca\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2 \u0d9a\u0dc5 \u0dba\u0dd4\u0dad\u0dd4 \u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0dda \u0db1\u0db8\u0dba\u0dd2 </li>\n<li><span translate=no>_^_1_^_</span> \u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8 \u0dc0\u0dda</li></ul>\n",
|
||||
"<h3>Clear a cache value</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name used when caching</li></ul>\n": "<h3>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0d85\u0d9c\u0dba \u0d89\u0dc0\u0dad\u0dca \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda\u0daf\u0dd3 \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db1 \u0db1\u0db8\u0dba\u0dd2</li></ul>\n",
|
||||
"<h3>Clear cache</h3>\n": "<h3>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba\u0d89\u0dc0\u0dad\u0dca \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n",
|
||||
"<h3>Get the cache instance</h3>\n<ul><p><em>Returns</em> the cache instance</p></ul>\n": "<h3>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0d8b\u0daf\u0dcf\u0dc4\u0dbb\u0dab\u0dba\u0d9a\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1</h3>\n<ul><p>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0d85\u0dc0\u0dc3\u0dca\u0dae\u0dcf\u0dc0<em>\u0d86\u0db4\u0dc3\u0dd4 \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<h3>Pop from a queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<p><em>Returns</em> the value</p></ul>\n": "<h3>\u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dda\u0dc3\u0dd2\u0da7 \u0db4\u0ddc\u0db4\u0dca</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dda \u0db1\u0db8 </li>\n<p>\u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8<em>\u0db1\u0dd0\u0dc0\u0dad \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<h3>Push a value to a queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<li><span translate=no>_^_1_^_</span> is the value to be pushed</li></ul>\n": "<h3>\u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0d9a\u0da7\u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0d9a\u0dca \u0dad\u0dbd\u0dca\u0dbd\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dda \u0db1\u0db8 </li>\n<li><span translate=no>_^_1_^_</span> \u0dad\u0dbd\u0dca\u0dbd\u0dd4 \u0d9a\u0dc5 \u0dba\u0dd4\u0dad\u0dd4 \u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8 \u0dc0\u0dda</li></ul>\n",
|
||||
"<h3>Retrieve a value from cache</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name used when caching </li>\n<li><span translate=no>_^_1_^_</span> is the default value if the cache is empty </li>\n<p><em>Returns</em> the cached value</p></ul>\n": "<h3>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dc0\u0dbd\u0dd2\u0db1\u0dca\u0dc0\u0da7\u0dd2\u0db1\u0dcf\u0d9a\u0db8\u0d9a\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda\u0daf\u0dd3 \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db1 \u0db1\u0db8\u0dba\u0dd2 </li>\n<li><span translate=no>_^_1_^_</span> \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba \u0dc4\u0dd2\u0dc3\u0dca \u0db1\u0db8\u0dca \u0db4\u0dd9\u0dbb\u0db1\u0dd2\u0db8\u0dd2 \u0d85\u0d9c\u0dba \u0dc0\u0dda </li>\n<p>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0d85\u0d9c\u0dba<em>\u0db1\u0dd0\u0dc0\u0dad \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<h3>Return the size of the queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<p><em>Returns</em> size of the queue if exists else None</p></ul>\n": "<h3>\u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dda\u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba \u0d86\u0db4\u0dc3\u0dd4 \u0daf\u0dd9\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dda \u0db1\u0db8 </li>\n</ul><p>\u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0dda<em>\u0db4\u0dca\u0dbb\u0dad\u0dd2\u0dbd\u0dcf\u0db7</em> \u0db4\u0dca\u0dbb\u0db8\u0dcf\u0dab\u0dba \u0dc0\u0dd9\u0db1\u0dad\u0dca \u0d9a\u0dd2\u0dc3\u0dd2\u0dc0\u0d9a\u0dca \u0dad\u0dd2\u0db6\u0dda \u0db1\u0db8\u0dca \u0db1\u0dd0\u0dad</p>\n",
|
||||
"<p>Create an empty queue if it's not present </p>\n": "<p>\u0d91\u0dba\u0db1\u0ddc\u0db8\u0dd0\u0dad\u0dd2 \u0db1\u0db8\u0dca \u0dc4\u0dd2\u0dc3\u0dca \u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0d9a\u0dca \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Push to the queue </p>\n": "<p>\u0db4\u0ddd\u0dbd\u0dd2\u0db8\u0da7\u0dad\u0dbd\u0dca\u0dbd\u0dd4 \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Singleton for cache </p>\n": "<p>\u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dc3\u0db3\u0dc4\u0dcf \u0dc3\u0dd2\u0d82\u0d9c\u0dbd\u0dca\u0da7\u0db1\u0dca </p>\n",
|
||||
"Cache for Intermediate Activations": "\u0d85\u0dad\u0dbb\u0db8\u0dd0\u0daf\u0dd2 \u0dc3\u0d9a\u0dca\u0dbb\u0dd3\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dca \u0dc3\u0db3\u0dc4\u0dcf \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2\u0dba",
|
||||
"Cache for intermediate activations for faster inference.": "\u0dc0\u0dda\u0d9c\u0dc0\u0dad\u0dca \u0d85\u0db1\u0dd4\u0db8\u0dcf\u0db1\u0dba \u0dc3\u0db3\u0dc4\u0dcf \u0d85\u0dad\u0dbb\u0db8\u0dd0\u0daf\u0dd2 \u0dc3\u0d9a\u0dca\u0dbb\u0dd3\u0dba \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dca \u0dc3\u0db3\u0dc4\u0dcf \u0dc4\u0dd0\u0db9\u0dd2\u0dbd\u0dd2."
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
{
|
||||
"<h1>Cache for Intermediate Activations</h1>\n<p>During inference the model outputs token by token. We use this simple cache to store key's and value's attention layers, so that we don't have to recompute them for previous tokens.</p>\n": "<h1>\u7528\u4e8e\u4e2d\u95f4\u6fc0\u6d3b\u7684\u7f13\u5b58</h1>\n<p>\u5728\u63a8\u7406\u8fc7\u7a0b\u4e2d\uff0c\u6a21\u578b\u9010\u4e2a\u8f93\u51fa\u4ee4\u724c\u3002\u6211\u4eec\u4f7f\u7528\u8fd9\u4e2a\u7b80\u5355\u7684\u7f13\u5b58\u6765\u5b58\u50a8\u952e\u548c\u503c\u7684\u6ce8\u610f\u5c42\uff0c\u8fd9\u6837\u6211\u4eec\u5c31\u4e0d\u5fc5\u4e3a\u4ee5\u524d\u7684\u4ee4\u724c\u91cd\u65b0\u8ba1\u7b97\u5b83\u4eec\u4e86\u3002</p>\n",
|
||||
"<h2>Cache</h2>\n<p>This maintains a key-value cache and queues push values and pop them in the same order. The queues are useful since we have multiple attention layers.</p>\n": "<h2>\u7f13\u5b58</h2>\n<p>\u8fd9\u5c06\u7ef4\u62a4\u4e00\u4e2a\u952e\u503c\u7f13\u5b58\uff0c\u5e76\u5c06\u63a8\u9001\u503c\u6392\u961f\u5e76\u6309\u76f8\u540c\u7684\u987a\u5e8f\u5f39\u51fa\u5b83\u4eec\u3002\u961f\u5217\u975e\u5e38\u6709\u7528\uff0c\u56e0\u4e3a\u6211\u4eec\u6709\u591a\u4e2a\u5173\u6ce8\u5c42\u3002</p>\n",
|
||||
"<h3>Cache a value</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the value to be cached </li>\n<li><span translate=no>_^_1_^_</span> is the value</li></ul>\n": "<h3>\u7f13\u5b58\u4e00\u4e2a\u503c</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u8981\u7f13\u5b58\u7684\u503c\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u4ef7\u503c</li></ul>\n",
|
||||
"<h3>Clear a cache value</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name used when caching</li></ul>\n": "<h3>\u6e05\u9664\u7f13\u5b58\u503c</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u7f13\u5b58\u65f6\u4f7f\u7528\u7684\u540d\u79f0</li></ul>\n",
|
||||
"<h3>Clear cache</h3>\n": "<h3>\u6e05\u9664\u7f13\u5b58</h3>\n",
|
||||
"<h3>Get the cache instance</h3>\n<ul><p><em>Returns</em> the cache instance</p></ul>\n": "<h3>\u83b7\u53d6\u7f13\u5b58\u5b9e\u4f8b</h3>\n<ul><p><em>\u8fd4\u56de</em>\u7f13\u5b58\u5b9e\u4f8b</p></ul>\n",
|
||||
"<h3>Pop from a queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<p><em>Returns</em> the value</p></ul>\n": "<h3>\u4ece\u961f\u5217\u4e2d\u5f39\u51fa</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u961f\u5217\u7684\u540d\u79f0</li>\n<p><em>\u8fd4\u56de</em>\u503c</p></ul>\n",
|
||||
"<h3>Push a value to a queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<li><span translate=no>_^_1_^_</span> is the value to be pushed</li></ul>\n": "<h3>\u5c06\u503c\u63a8\u9001\u5230\u961f\u5217</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u961f\u5217\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u8981\u63a8\u9001\u7684\u503c</li></ul>\n",
|
||||
"<h3>Retrieve a value from cache</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name used when caching </li>\n<li><span translate=no>_^_1_^_</span> is the default value if the cache is empty </li>\n<p><em>Returns</em> the cached value</p></ul>\n": "<h3>\u4ece\u7f13\u5b58\u4e2d\u68c0\u7d22\u503c</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u7f13\u5b58\u65f6\u4f7f\u7528\u7684\u540d\u79f0</li>\n<li><span translate=no>_^_1_^_</span>\u5982\u679c\u7f13\u5b58\u4e3a\u7a7a\uff0c\u5219\u4e3a\u9ed8\u8ba4\u503c</li>\n<p><em>\u8fd4\u56de</em>\u7f13\u5b58\u7684\u503c</p></ul>\n",
|
||||
"<h3>Return the size of the queue</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the name of the queue </li>\n<p><em>Returns</em> size of the queue if exists else None</p></ul>\n": "<h3>\u8fd4\u56de\u961f\u5217\u7684\u5927\u5c0f</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u961f\u5217\u7684\u540d\u79f0</li>\n<p><em>\u8fd4\u56de</em>\u961f\u5217\u7684\u5927\u5c0f\uff08\u5982\u679c\u5b58\u5728\uff09\u5426\u5219 None</p></ul>\n",
|
||||
"<p>Create an empty queue if it's not present </p>\n": "<p>\u5982\u679c\u961f\u5217\u4e0d\u5b58\u5728\uff0c\u8bf7\u521b\u5efa\u4e00\u4e2a\u7a7a\u961f\u5217</p>\n",
|
||||
"<p>Push to the queue </p>\n": "<p>\u63a8\u9001\u5230\u961f\u5217</p>\n",
|
||||
"<p>Singleton for cache </p>\n": "<p>\u7f13\u5b58\u7684\u5355\u4f8b</p>\n",
|
||||
"Cache for Intermediate Activations": "\u7528\u4e8e\u4e2d\u95f4\u6fc0\u6d3b\u7684\u7f13\u5b58",
|
||||
"Cache for intermediate activations for faster inference.": "\u7f13\u5b58\u7528\u4e8e\u4e2d\u95f4\u6fc0\u6d3b\uff0c\u4ee5\u4fbf\u66f4\u5feb\u5730\u63a8\u65ad\u3002"
|
||||
}
|
||||
@@ -0,0 +1,6 @@
|
||||
{
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>No need to train the mlp bias because we are adding it with attention output </p>\n": "<p>\u6ce8\u610f\u51fa\u529b\u3067\u52a0\u7b97\u3059\u308b\u306e\u3067\u3001mlp\u30d0\u30a4\u30a2\u30b9\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u5fc5\u8981\u306f\u3042\u308a\u307e\u305b\u3093\u3002</p>\n",
|
||||
"<p>Set <span translate=no>_^_0_^_</span> to <span translate=no>_^_1_^_</span> for the entire layer. </p>\n": "<p><span translate=no>_^_0_^_</span><span translate=no>_^_1_^_</span>\u30ec\u30a4\u30e4\u30fc\u5168\u4f53\u3067\u306b\u8a2d\u5b9a\u3057\u307e\u3059\u3002</p>\n",
|
||||
"finetune.py": "finetune.py"
|
||||
}
|
||||
@@ -0,0 +1,6 @@
|
||||
{
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>No need to train the mlp bias because we are adding it with attention output </p>\n": "<p>mlp\u0db1\u0dd0\u0db9\u0dd4\u0dbb\u0dd4\u0dc0 \u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0d85\u0dc0\u0dc1\u0dca\u0dba \u0db1\u0ddc\u0dc0\u0dda, \u0db8\u0db1\u0dca\u0daf \u0d85\u0db4\u0dd2 \u0d91\u0dba \u0d85\u0dc0\u0db0\u0dcf\u0db1\u0dba \u0dba\u0ddc\u0db8\u0dd4 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda \u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba \u0dc3\u0db8\u0d9f \u0d91\u0d9a\u0dad\u0dd4 \u0d9a\u0dbb\u0db1 \u0db6\u0dd0\u0dc0\u0dd2\u0db1\u0dd2 </p>\n",
|
||||
"<p>Set <span translate=no>_^_0_^_</span> to <span translate=no>_^_1_^_</span> for the entire layer. </p>\n": "<p>\u0dc3\u0db8\u0dca\u0db4\u0dd6\u0dbb\u0dca\u0dab\u0dc3\u0dca\u0dad\u0dbb\u0dba <span translate=no>_^_0_^_</span> <span translate=no>_^_1_^_</span> \u0dc3\u0db3\u0dc4\u0dcf \u0dc3\u0d9a\u0dc3\u0db1\u0dca\u0db1. </p>\n",
|
||||
"finetune.py": "finetune.py"
|
||||
}
|
||||
@@ -0,0 +1,6 @@
|
||||
{
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>No need to train the mlp bias because we are adding it with attention output </p>\n": "<p>\u4e0d\u9700\u8981\u8bad\u7ec3 mlp \u504f\u5dee\uff0c\u56e0\u4e3a\u6211\u4eec\u6dfb\u52a0\u4e86\u6ce8\u610f\u529b\u8f93\u51fa</p>\n",
|
||||
"<p>Set <span translate=no>_^_0_^_</span> to <span translate=no>_^_1_^_</span> for the entire layer. </p>\n": "<p><span translate=no>_^_0_^_</span>\u5c06<span translate=no>_^_1_^_</span>\u6574\u4e2a\u56fe\u5c42\u8bbe\u7f6e\u4e3a\u3002</p>\n",
|
||||
"finetune.py": "finetune.py"
|
||||
}
|
||||
@@ -0,0 +1,11 @@
|
||||
{
|
||||
"<h1>LLM.int() on GPT-NeoX</h1>\n<p>This implements a utility function to transform a <span translate=no>_^_0_^_</span> layer to LLM.int8() linear layer.</p>\n<p><a href=\"https://arxiv.org/abs/eb2bcaee1d0011edaa66a71c10a887e7\">LLM.int8() paper</a> shows you can use int8 quantization while handling outliers to reduce memory footprint without performance degradation in large language models. They convert weights and inputs to scaled 8-bit integers and does matrix multiplication producing int32 results which is then converted back to float16 and rescaled. They show that in large langauge models, some features can give extreme values (outliers) that dominate the model's output. These features get clamped in 8-bit integer space which causes the model performance to degrade. As a solution they pick these outliers (greater than a specified threshold) and compute their multiplications separately in float16 space. Since the percentage of outliers is around 0.01% this doesn't increase memory usage, and prevents the model from degrading performance.</p>\n<p>The code to transform GPT-NoeX layers is defined in <a href=\"../model.html#post_load_prepare\">model.py</a>.</p>\n<p>Here are example uses of GPT-NeoX with int8 quantization.</p>\n<ul><li><a href=\"../samples/llm_int8.html\">Generate Text</a> </li>\n<li><a href=\"../evaluation/llm_int8.html\">Run Evaluation Tests</a></li></ul>\n": "<h1>GPT \u30cd\u30aa\u30c3\u30af\u30b9\u306e llm.int ()</h1>\n<p>\u3053\u308c\u306f\u3001\u30ec\u30a4\u30e4\u30fc\u3092 llm.int8 () <span translate=no>_^_0_^_</span> \u7dda\u5f62\u30ec\u30a4\u30e4\u30fc\u306b\u5909\u63db\u3059\u308b\u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3\u95a2\u6570\u3092\u5b9f\u88c5\u3057\u307e\u3059\u3002</p>\n<p><a href=\"https://arxiv.org/abs/eb2bcaee1d0011edaa66a71c10a887e7\">LLM.int8 () \u306e\u8ad6\u6587\u3067\u306f</a>\u3001\u5927\u898f\u6a21\u306a\u8a00\u8a9e\u30e2\u30c7\u30eb\u3067\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u4f4e\u4e0b\u3055\u305b\u308b\u3053\u3068\u306a\u304f\u3001\u5916\u308c\u5024\u3092\u51e6\u7406\u3059\u308b\u969b\u306b int8 \u91cf\u5b50\u5316\u3092\u4f7f\u7528\u3057\u3066\u30e1\u30e2\u30ea\u4f7f\u7528\u91cf\u3092\u524a\u6e1b\u3067\u304d\u308b\u3053\u3068\u304c\u793a\u3055\u308c\u3066\u3044\u307e\u3059\u3002\u91cd\u307f\u3068\u5165\u529b\u3092\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3055\u308c\u305f8\u30d3\u30c3\u30c8\u6574\u6570\u306b\u5909\u63db\u3057\u3001\u884c\u5217\u4e57\u7b97\u3092\u884c\u3063\u3066int32\u306e\u7d50\u679c\u3092\u751f\u6210\u3057\u3001\u305d\u308c\u3092float16\u306b\u5909\u63db\u3057\u3066\u518d\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3057\u307e\u3059\u3002\u305d\u306e\u7d50\u679c\u3001\u5927\u898f\u6a21\u306a\u8a00\u8a9e\u30e2\u30c7\u30eb\u3067\u306f\u3001\u4e00\u90e8\u306e\u7279\u5fb4\u306b\u3088\u3063\u3066\u6975\u7aef\u306a\u5024 (\u5916\u308c\u5024) \u304c\u30e2\u30c7\u30eb\u306e\u51fa\u529b\u306e\u5927\u90e8\u5206\u3092\u5360\u3081\u308b\u53ef\u80fd\u6027\u304c\u3042\u308b\u3053\u3068\u304c\u793a\u3055\u308c\u3066\u3044\u307e\u3059\u3002\u3053\u308c\u3089\u306e\u6a5f\u80fd\u306f8\u30d3\u30c3\u30c8\u306e\u6574\u6570\u7a7a\u9593\u306b\u5236\u9650\u3055\u308c\u308b\u305f\u3081\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u304c\u4f4e\u4e0b\u3057\u307e\u3059\u3002\u89e3\u6c7a\u7b56\u3068\u3057\u3066\u3001\u3053\u308c\u3089\u306e\u5916\u308c\u5024\uff08\u6307\u5b9a\u3055\u308c\u305f\u3057\u304d\u3044\u5024\u3092\u8d85\u3048\u308b\uff09\u3092\u9078\u629e\u3057\u3001\u305d\u306e\u4e57\u7b97\u3092float16\u7a7a\u9593\u3067\u500b\u5225\u306b\u8a08\u7b97\u3057\u307e\u3059\u3002\u5916\u308c\u5024\u306e\u5272\u5408\u306f\u7d04 0.01% \u306a\u306e\u3067\u3001\u30e1\u30e2\u30ea\u4f7f\u7528\u91cf\u306f\u5897\u52a0\u305b\u305a\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u4f4e\u4e0b\u3092\u9632\u304e\u307e\u3059</p>\u3002\n<p><a href=\"../model.html#post_load_prepare\">GPT-NoEx \u30ec\u30a4\u30e4\u30fc\u3092\u5909\u63db\u3059\u308b\u30b3\u30fc\u30c9\u306f model.py \u3067\u5b9a\u7fa9\u3055\u308c\u3066\u3044\u307e\u3059\u3002</a></p>\n<p>int8 \u91cf\u5b50\u5316\u306b\u3088\u308b GPT-Neox \u306e\u4f7f\u7528\u4f8b\u3092\u4ee5\u4e0b\u306b\u793a\u3057\u307e\u3059\u3002</p>\n<ul><li><a href=\"../samples/llm_int8.html\">\u30c6\u30ad\u30b9\u30c8\u3092\u751f\u6210</a></li>\n<li><a href=\"../evaluation/llm_int8.html\">\u8a55\u4fa1\u30c6\u30b9\u30c8\u3092\u5b9f\u884c</a></li></ul>\n",
|
||||
"<h2>Transform a <span translate=no>_^_0_^_</span> layer to LLM.int8() linear layer</h2>\n<ul><li><span translate=no>_^_1_^_</span> is the <span translate=no>_^_2_^_</span> layer to transform </li>\n<li><span translate=no>_^_3_^_</span> is the device of the model </li>\n<li><span translate=no>_^_4_^_</span> is the threshold <span translate=no>_^_5_^_</span> to use for outlier detection</li></ul>\n": "<h2><span translate=no>_^_0_^_</span>\u30ec\u30a4\u30e4\u30fc\u3092 llm.int8 () \u30ea\u30cb\u30a2\u30ec\u30a4\u30e4\u30fc\u306b\u5909\u63db\u3057\u307e\u3059</h2>\n<ul><li><span translate=no>_^_1_^_</span><span translate=no>_^_2_^_</span>\u5909\u63db\u3059\u308b\u30ec\u30a4\u30e4\u30fc\u3067\u3059</li>\n<li><span translate=no>_^_3_^_</span>\u30e2\u30c7\u30eb\u306e\u30c7\u30d0\u30a4\u30b9\u3067\u3059</li>\n<li><span translate=no>_^_4_^_</span><span translate=no>_^_5_^_</span>\u5916\u308c\u5024\u306e\u691c\u51fa\u306b\u4f7f\u7528\u3059\u308b\u95be\u5024\u3067\u3059</li></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create an empty Linear8bitLt module </p>\n": "<p>\u7a7a\u306e Linear8BitLT \u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002</p>\n",
|
||||
"<p>Import <a href=\"https://github.com/timdettmers/bitsandbytes\"><span translate=no>_^_0_^_</span></a> package </p>\n": "<p><a href=\"https://github.com/timdettmers/bitsandbytes\"><span translate=no>_^_0_^_</span></a>\u30d1\u30c3\u30b1\u30fc\u30b8\u3092\u30a4\u30f3\u30dd\u30fc\u30c8</p>\n",
|
||||
"<p>Quantize the weights </p>\n": "<p>\u30a6\u30a7\u30a4\u30c8\u3092\u30af\u30aa\u30f3\u30bf\u30a4\u30ba</p>\n",
|
||||
"<p>Set the bias in float16 space </p>\n": "<p>float16 \u7a7a\u9593\u306b\u30d0\u30a4\u30a2\u30b9\u3092\u8a2d\u5b9a\u3057\u307e\u3059\u3002</p>\n",
|
||||
"LLM.int8() on GPT-NeoX": "GPT \u30cd\u30aa\u30c3\u30af\u30b9\u306e llm.int8 ()",
|
||||
"Transform nn.Linear layers to 8-bit integer layers.": "nn.linear \u5c64\u3092 8 \u30d3\u30c3\u30c8\u306e\u6574\u6570\u5c64\u306b\u5909\u63db\u3057\u307e\u3059\u3002"
|
||||
}
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,11 @@
|
||||
{
|
||||
"<h1>LLM.int() on GPT-NeoX</h1>\n<p>This implements a utility function to transform a <span translate=no>_^_0_^_</span> layer to LLM.int8() linear layer.</p>\n<p><a href=\"https://arxiv.org/abs/eb2bcaee1d0011edaa66a71c10a887e7\">LLM.int8() paper</a> shows you can use int8 quantization while handling outliers to reduce memory footprint without performance degradation in large language models. They convert weights and inputs to scaled 8-bit integers and does matrix multiplication producing int32 results which is then converted back to float16 and rescaled. They show that in large langauge models, some features can give extreme values (outliers) that dominate the model's output. These features get clamped in 8-bit integer space which causes the model performance to degrade. As a solution they pick these outliers (greater than a specified threshold) and compute their multiplications separately in float16 space. Since the percentage of outliers is around 0.01% this doesn't increase memory usage, and prevents the model from degrading performance.</p>\n<p>The code to transform GPT-NoeX layers is defined in <a href=\"../model.html#post_load_prepare\">model.py</a>.</p>\n<p>Here are example uses of GPT-NeoX with int8 quantization.</p>\n<ul><li><a href=\"../samples/llm_int8.html\">Generate Text</a> </li>\n<li><a href=\"../evaluation/llm_int8.html\">Run Evaluation Tests</a></li></ul>\n": "<h1>GPT-NEOX \u4e0a\u7684 llm.int ()</h1>\n<p>\u8fd9\u5b9e\u73b0\u4e86\u4e00\u4e2a\u5b9e\u7528\u7a0b\u5e8f\u51fd\u6570\uff0c\u5c06<span translate=no>_^_0_^_</span>\u5c42\u8f6c\u6362\u4e3a LLM.int8 () \u7ebf\u6027\u5c42\u3002</p>\n<p><a href=\"https://arxiv.org/abs/eb2bcaee1d0011edaa66a71c10a887e7\">LLM.int8 () \u8bba\u6587</a>\u5c55\u793a\u4e86\u5728\u5904\u7406\u5f02\u5e38\u503c\u65f6\u53ef\u4ee5\u4f7f\u7528 int8 \u91cf\u5316\u6765\u51cf\u5c11\u5185\u5b58\u5360\u7528\uff0c\u800c\u4e0d\u4f1a\u964d\u4f4e\u5927\u578b\u8bed\u8a00\u6a21\u578b\u7684\u6027\u80fd\u3002\u5b83\u4eec\u5c06\u6743\u91cd\u548c\u8f93\u5165\u8f6c\u6362\u4e3a\u6309\u6bd4\u4f8b\u7f29\u653e\u76848\u4f4d\u6574\u6570\uff0c\u5e76\u8fdb\u884c\u77e9\u9635\u4e58\u6cd5\u4ea7\u751fint32\u7ed3\u679c\uff0c\u7136\u540e\u5c06\u5176\u8f6c\u6362\u56defloat16\u5e76\u91cd\u65b0\u7f29\u653e\u3002\u5b83\u4eec\u8868\u660e\uff0c\u5728\u5927\u578b\u8bed\u8a00\u6a21\u578b\u4e2d\uff0c\u67d0\u4e9b\u7279\u5f81\u53ef\u4ee5\u7ed9\u51fa\u6781\u503c\uff08\u5f02\u5e38\u503c\uff09\uff0c\u8fd9\u4e9b\u503c\u5728\u6a21\u578b\u7684\u8f93\u51fa\u4e2d\u5360\u636e\u4e3b\u5bfc\u5730\u4f4d\u3002\u8fd9\u4e9b\u7279\u5f81\u88ab\u9650\u5236\u5728 8 \u4f4d\u6574\u6570\u7a7a\u95f4\u4e2d\uff0c\u8fd9\u4f1a\u5bfc\u81f4\u6a21\u578b\u6027\u80fd\u4e0b\u964d\u3002\u4f5c\u4e3a\u89e3\u51b3\u65b9\u6848\uff0c\u4ed6\u4eec\u9009\u62e9\u8fd9\u4e9b\u5f02\u5e38\u503c\uff08\u5927\u4e8e\u6307\u5b9a\u9608\u503c\uff09\uff0c\u5e76\u5728float16\u7a7a\u95f4\u4e2d\u5206\u522b\u8ba1\u7b97\u5b83\u4eec\u7684\u4e58\u6cd5\u3002\u7531\u4e8e\u5f02\u5e38\u503c\u7684\u767e\u5206\u6bd4\u7ea6\u4e3a 0.01%\uff0c\u56e0\u6b64\u4e0d\u4f1a\u589e\u52a0\u5185\u5b58\u4f7f\u7528\u91cf\uff0c\u5e76\u9632\u6b62\u6a21\u578b\u964d\u4f4e\u6027\u80fd\u3002</p>\n<p>\u7528\u4e8e\u8f6c\u6362 GPT-NOEX \u5c42\u7684\u4ee3\u7801\u5728 <a href=\"../model.html#post_load_prepare\">model.py</a> \u4e2d\u5b9a\u4e49\u3002</p>\n<p>\u4ee5\u4e0b\u662f\u4f7f\u7528 int8 \u91cf\u5316\u7684 GPT-NEOX \u7684\u793a\u4f8b\u7528\u6cd5\u3002</p>\n<ul><li><a href=\"../samples/llm_int8.html\">\u751f\u6210\u6587\u672c</a></li>\n<li><a href=\"../evaluation/llm_int8.html\">\u8fd0\u884c\u8bc4\u4f30\u6d4b\u8bd5</a></li></ul>\n",
|
||||
"<h2>Transform a <span translate=no>_^_0_^_</span> layer to LLM.int8() linear layer</h2>\n<ul><li><span translate=no>_^_1_^_</span> is the <span translate=no>_^_2_^_</span> layer to transform </li>\n<li><span translate=no>_^_3_^_</span> is the device of the model </li>\n<li><span translate=no>_^_4_^_</span> is the threshold <span translate=no>_^_5_^_</span> to use for outlier detection</li></ul>\n": "<h2>\u5c06<span translate=no>_^_0_^_</span>\u56fe\u5c42\u8f6c\u6362\u4e3a LLM.int8 () \u7ebf\u6027\u56fe\u5c42</h2>\n<ul><li><span translate=no>_^_1_^_</span>\u662f\u8981\u53d8\u6362\u7684<span translate=no>_^_2_^_</span>\u56fe\u5c42</li>\n<li><span translate=no>_^_3_^_</span>\u662f\u8be5\u578b\u53f7\u7684\u8bbe\u5907</li>\n<li><span translate=no>_^_4_^_</span>\u662f\u7528\u4e8e\u5f02\u5e38\u503c\u68c0\u6d4b\u7684\u9608<span translate=no>_^_5_^_</span>\u503c</li></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create an empty Linear8bitLt module </p>\n": "<p>\u521b\u5efa\u4e00\u4e2a\u7a7a\u7684 Linear8bitLT \u6a21\u5757</p>\n",
|
||||
"<p>Import <a href=\"https://github.com/timdettmers/bitsandbytes\"><span translate=no>_^_0_^_</span></a> package </p>\n": "<p>\u5bfc\u5165<a href=\"https://github.com/timdettmers/bitsandbytes\"><span translate=no>_^_0_^_</span></a>\u5305</p>\n",
|
||||
"<p>Quantize the weights </p>\n": "<p>\u91cf\u5316\u6743\u91cd</p>\n",
|
||||
"<p>Set the bias in float16 space </p>\n": "<p>\u5728 float16 \u7a7a\u95f4\u4e2d\u8bbe\u7f6e\u504f\u5dee</p>\n",
|
||||
"LLM.int8() on GPT-NeoX": "GPT-NEOX \u4e0a\u7684 llm.int8 ()",
|
||||
"Transform nn.Linear layers to 8-bit integer layers.": "\u5c06 NN. \u7ebf\u6027\u56fe\u5c42\u8f6c\u6362\u4e3a 8 \u4f4d\u6574\u6570\u56fe\u5c42\u3002"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h1>Text Dataset for GPT-NeoX</h1>\n": "<h1>GPT-Neox \u7528\u30c6\u30ad\u30b9\u30c8\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8</h1>\n",
|
||||
"<h2>Dataset for fine-tuning GPT-NeoX</h2>\n<p>This is not optimized to very large datasets.</p>\n": "<h2>GPT-Neox \u3092\u5fae\u8abf\u6574\u3059\u308b\u305f\u3081\u306e\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8</h2>\n<p>\u3053\u308c\u306f\u975e\u5e38\u306b\u5927\u304d\u306a\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306b\u306f\u6700\u9069\u5316\u3055\u308c\u3066\u3044\u307e\u305b\u3093\u3002</p>\n",
|
||||
"<h3>Get a sample</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the index of the sample </li>\n<p><em>Returns</em> the input and the target</p></ul>\n": "<h3>\u30b5\u30f3\u30d7\u30eb\u3092\u5165\u624b</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30b5\u30f3\u30d7\u30eb\u306e\u30a4\u30f3\u30c7\u30c3\u30af\u30b9\u3067\u3059</li>\n<p><em>\u5165\u529b\u3068\u30bf\u30fc\u30b2\u30c3\u30c8\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h3>Load Dataset</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the sequence length of a single training sample </li>\n<li><span translate=no>_^_1_^_</span> is the name of the dataset </li>\n<p><em>Returns</em> the dataset</p></ul>\n": "<h3>\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u8aad\u307f\u8fbc\u307f</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u306f 1 \u3064\u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30b5\u30f3\u30d7\u30eb\u306e\u30b7\u30fc\u30b1\u30f3\u30b9\u9577\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u540d\u524d</li>\n<p><em>\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<h3>Load text file</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the location of the text file </li>\n<li><span translate=no>_^_1_^_</span> is the URL to download the file from </li>\n<li><span translate=no>_^_2_^_</span> is the number of characters to filter. Use this during testing when trying large datasets </li>\n<p><em>Returns</em> the text content</p></ul>\n": "<h3>\u30c6\u30ad\u30b9\u30c8\u30d5\u30a1\u30a4\u30eb\u306e\u8aad\u307f\u8fbc\u307f</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30c6\u30ad\u30b9\u30c8\u30d5\u30a1\u30a4\u30eb\u306e\u5834\u6240\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u30d5\u30a1\u30a4\u30eb\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3059\u308b URL</li>\n<li><span translate=no>_^_2_^_</span>\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u3059\u308b\u6587\u5b57\u306e\u6570\u3067\u3059\u3002\u5927\u898f\u6a21\u306a\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u8a66\u3059\u3068\u304d\u306e\u30c6\u30b9\u30c8\u6642\u306b\u3053\u308c\u3092\u4f7f\u7528\u3057\u3066\u304f\u3060\u3055\u3044</li>\n<p><em>\u30c6\u30ad\u30b9\u30c8\u30b3\u30f3\u30c6\u30f3\u30c4\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create a PyTorch tensor </p>\n": "<p>PyTorch \u30c6\u30f3\u30bd\u30eb\u306e\u4f5c\u6210</p>\n",
|
||||
"<p>Download if it doesn't exist </p>\n": "<p>\u5b58\u5728\u3057\u306a\u3044\u5834\u5408\u306f\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9</p>\n",
|
||||
"<p>Filter </p>\n": "<p>\u30d5\u30a3\u30eb\u30bf</p>\n",
|
||||
"<p>Load data </p>\n": "<p>\u30c7\u30fc\u30bf\u3092\u8aad\u307f\u8fbc\u3080</p>\n",
|
||||
"<p>Load the content </p>\n": "<p>\u30b3\u30f3\u30c6\u30f3\u30c4\u3092\u8aad\u307f\u8fbc\u3080</p>\n",
|
||||
"<p>Number of samples </p>\n": "<p>\u30b5\u30f3\u30d7\u30eb\u6570</p>\n",
|
||||
"<p>Tokenize </p>\n": "<p>\u30c8\u30fc\u30af\u30f3\u5316</p>\n",
|
||||
"<p>Truncate </p>\n": "<p>\u5207\u308a\u6368\u3066</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the list of token ids </li>\n<li><span translate=no>_^_1_^_</span> is the sequence length of a single training sample</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u30c8\u30fc\u30af\u30f3 ID \u306e\u30ea\u30b9\u30c8\u3067\u3059</li>\n<li><span translate=no>_^_1_^_</span>\u306f 1 \u3064\u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30b5\u30f3\u30d7\u30eb\u306e\u30b7\u30fc\u30b1\u30f3\u30b9\u9577\u3067\u3059</li></ul>\n",
|
||||
"Loads text datasets to fine-tune GPT-NeoX": "\u30c6\u30ad\u30b9\u30c8\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u30ed\u30fc\u30c9\u3057\u3066GPT-Neox\u3092\u5fae\u8abf\u6574\u3057\u307e\u3059",
|
||||
"Text Dataset for GPT-NeoX": "GPT-Neox \u7528\u30c6\u30ad\u30b9\u30c8\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h1>Text Dataset for GPT-NeoX</h1>\n": "<h1>\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca\u0dc3\u0db3\u0dc4\u0dcf \u0db4\u0dd9\u0dc5 \u0daf\u0dad\u0dca\u0dad \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba</h1>\n",
|
||||
"<h2>Dataset for fine-tuning GPT-NeoX</h2>\n<p>This is not optimized to very large datasets.</p>\n": "<h2>\u0dc4\u0ddc\u0db3\u0dd2\u0db1\u0dca\u0dc3\u0dd4\u0dc3\u0dbb \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0daf\u0dad\u0dca\u0dad \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba GPT-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca</h2>\n<p>\u0db8\u0dd9\u0dba\u0d89\u0dad\u0dcf \u0dc0\u0dd2\u0dc1\u0dcf\u0dbd \u0daf\u0dad\u0dca\u0dad \u0d9a\u0dcf\u0dab\u0dca\u0da9 \u0dc0\u0dbd\u0da7 \u0db4\u0dca\u0dbb\u0dc1\u0dc3\u0dca\u0dad\u0dd2\u0d9a\u0dbb\u0dab\u0dba \u0d9a\u0dbb \u0db1\u0ddc\u0db8\u0dd0\u0dad. </p>\n",
|
||||
"<h3>Get a sample</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the index of the sample </li>\n<p><em>Returns</em> the input and the target</p></ul>\n": "<h3>\u0db1\u0dd2\u0dba\u0dd0\u0daf\u0dd2\u0dba\u0d9a\u0dca\u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db1\u0dd2\u0dba\u0dd0\u0daf\u0dd2\u0dba\u0dda \u0daf\u0dbb\u0dca\u0dc1\u0d9a\u0dba \u0dc0\u0dda </li>\n<p>\u0d86\u0daf\u0dcf\u0db1\u0dba\u0dc3\u0dc4 \u0d89\u0dbd\u0d9a\u0dca\u0d9a\u0dba<em>\u0d86\u0db4\u0dc3\u0dd4 \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<h3>Load Dataset</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the sequence length of a single training sample </li>\n<li><span translate=no>_^_1_^_</span> is the name of the dataset </li>\n<p><em>Returns</em> the dataset</p></ul>\n": "<h3>\u0daf\u0dad\u0dca\u0dad\u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba \u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0dad\u0db1\u0dd2 \u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4 \u0db1\u0dd2\u0dba\u0dd0\u0daf\u0dd2\u0dba\u0d9a \u0d85\u0db1\u0dd4\u0d9a\u0dca\u0dbb\u0db8\u0dba \u0daf\u0dd2\u0d9c \u0dc0\u0dda </li>\n<li><span translate=no>_^_1_^_</span> \u0daf\u0dad\u0dca\u0dad \u0dc3\u0db8\u0dd4\u0daf\u0dcf\u0dba \u0db1\u0db8 </li>\n<p>\u0daf\u0dad\u0dca\u0dad\u0dc3\u0db8\u0dd4\u0daf\u0dcf\u0dba<em>\u0d86\u0db4\u0dc3\u0dd4 \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<h3>Load text file</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the location of the text file </li>\n<li><span translate=no>_^_1_^_</span> is the URL to download the file from </li>\n<li><span translate=no>_^_2_^_</span> is the number of characters to filter. Use this during testing when trying large datasets </li>\n<p><em>Returns</em> the text content</p></ul>\n": "<h3>\u0db4\u0dd9\u0dc5\u0d9c\u0ddc\u0db1\u0dd4\u0dc0 \u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0dd9\u0dc5 \u0d9c\u0ddc\u0db1\u0dd4\u0dc0\u0dda \u0db4\u0dd2\u0dc4\u0dd2\u0da7\u0dd3\u0db8 </li>\n<li><span translate=no>_^_1_^_</span> \u0d9c\u0ddc\u0db1\u0dd4\u0dc0 \u0db6\u0dcf\u0d9c\u0dad \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf URL \u0d91\u0d9a </li>\n<li><span translate=no>_^_2_^_</span> \u0dba\u0db1\u0dd4 \u0db4\u0dd9\u0dbb\u0dd3\u0db8\u0da7 \u0d85\u0d9a\u0dca\u0dc2\u0dbb \u0d9c\u0dab\u0db1\u0dba\u0dba\u0dd2. \u0dc0\u0dd2\u0dc1\u0dcf\u0dbd \u0daf\u0dad\u0dca\u0dad \u0d9a\u0da7\u0dca\u0da7\u0dbd \u0d8b\u0dad\u0dca\u0dc3\u0dcf\u0dc4 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda\u0daf\u0dd3 \u0db4\u0dbb\u0dd3\u0d9a\u0dca\u0dc2\u0dcf \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda\u0daf\u0dd3 \u0db8\u0dd9\u0dba \u0db7\u0dcf\u0dc0\u0dd2\u0dad\u0dcf \u0d9a\u0dbb\u0db1\u0dca\u0db1 </li>\n<p>\u0db4\u0dd9\u0dc5\u0d85\u0db1\u0dca\u0dad\u0dbb\u0dca\u0d9c\u0dad\u0dba<em>\u0db1\u0dd0\u0dc0\u0dad \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Create a PyTorch tensor </p>\n": "<p>\u0db4\u0dba\u0dd2\u0da7\u0ddd\u0da0\u0dca\u0da7\u0dd9\u0db1\u0dca\u0dc3\u0dbb\u0dba\u0d9a\u0dca \u0dc3\u0dcf\u0daf\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Download if it doesn't exist </p>\n": "<p>\u0d91\u0dba\u0db1\u0ddc\u0db8\u0dd0\u0dad\u0dd2 \u0db1\u0db8\u0dca \u0db6\u0dcf\u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Filter </p>\n": "<p>\u0db4\u0dd9\u0dbb\u0dc4\u0db1\u0dca </p>\n",
|
||||
"<p>Load data </p>\n": "<p>\u0daf\u0dad\u0dca\u0dad\u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Load the content </p>\n": "<p>\u0d85\u0db1\u0dca\u0dad\u0dbb\u0dca\u0d9c\u0dad\u0dba\u0db4\u0dd6\u0dbb\u0dab\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Number of samples </p>\n": "<p>\u0dc3\u0dcf\u0db8\u0dca\u0db4\u0dbd\u0d9c\u0dab\u0db1 </p>\n",
|
||||
"<p>Tokenize </p>\n": "<p>\u0da7\u0ddd\u0d9a\u0db1\u0dba\u0dd2\u0dc3\u0dca\u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Truncate </p>\n": "<p>\u0da7\u0dca\u0dbb\u0db1\u0dca\u0d9a\u0dda\u0da7\u0dca\u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the list of token ids </li>\n<li><span translate=no>_^_1_^_</span> is the sequence length of a single training sample</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span> \u0da7\u0ddd\u0d9a\u0db1\u0dca \u0dc4\u0dd0\u0db3\u0dd4\u0db1\u0dd4\u0db8\u0dca\u0db4\u0dad\u0dca \u0dbd\u0dd0\u0dba\u0dd2\u0dc3\u0dca\u0dad\u0dd4\u0dc0\u0dba\u0dd2 </li>\n<li><span translate=no>_^_1_^_</span> \u0dad\u0db1\u0dd2 \u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4 \u0db1\u0dd2\u0dba\u0dd0\u0daf\u0dd2\u0dba\u0d9a \u0d85\u0db1\u0dd4\u0d9a\u0dca\u0dbb\u0db8\u0dba \u0daf\u0dd2\u0d9c \u0dc0\u0dda</li></ul>\n",
|
||||
"Loads text datasets to fine-tune GPT-NeoX": "GPT-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0db8\u0db1\u0dcf\u0dc0 \u0dc3\u0d9a\u0dc3\u0dca \u0d9a\u0dd2\u0dbb\u0dd3\u0db8 \u0dc3\u0db3\u0dc4\u0dcf \u0db4\u0dd9\u0dc5 \u0daf\u0dad\u0dca\u0dad \u0d9a\u0da7\u0dca\u0da7\u0dbd \u0db4\u0da7\u0dc0\u0dba\u0dd2",
|
||||
"Text Dataset for GPT-NeoX": "\u0da2\u0dd3\u0db4\u0dd3\u0da7\u0dd3-\u0db1\u0dd2\u0dba\u0ddd\u0d9a\u0dca\u0dc3\u0dca \u0dc3\u0db3\u0dc4\u0dcf \u0db4\u0dd9\u0dc5 \u0daf\u0dad\u0dca\u0dad \u0d9a\u0da7\u0dca\u0da7\u0dbd\u0dba"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h1>Text Dataset for GPT-NeoX</h1>\n": "<h1>GPT-NEOX \u7684\u6587\u672c\u6570\u636e\u96c6</h1>\n",
|
||||
"<h2>Dataset for fine-tuning GPT-NeoX</h2>\n<p>This is not optimized to very large datasets.</p>\n": "<h2>\u7528\u4e8e\u5fae\u8c03 GPT-NEOX \u7684\u6570\u636e\u96c6</h2>\n<p>\u8fd9\u5e76\u672a\u9488\u5bf9\u975e\u5e38\u5927\u7684\u6570\u636e\u96c6\u8fdb\u884c\u4f18\u5316\u3002</p>\n",
|
||||
"<h3>Get a sample</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the index of the sample </li>\n<p><em>Returns</em> the input and the target</p></ul>\n": "<h3>\u83b7\u53d6\u6837\u54c1</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u6837\u672c\u7684\u7d22\u5f15</li>\n<p><em>\u8fd4\u56de</em>\u8f93\u5165\u548c\u76ee\u6807</p></ul>\n",
|
||||
"<h3>Load Dataset</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the sequence length of a single training sample </li>\n<li><span translate=no>_^_1_^_</span> is the name of the dataset </li>\n<p><em>Returns</em> the dataset</p></ul>\n": "<h3>\u52a0\u8f7d\u6570\u636e\u96c6</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u5355\u4e2a\u8bad\u7ec3\u6837\u672c\u7684\u5e8f\u5217\u957f\u5ea6</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u6570\u636e\u96c6\u7684\u540d\u79f0</li>\n<p><em>\u8fd4\u56de</em>\u6570\u636e\u96c6</p></ul>\n",
|
||||
"<h3>Load text file</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the location of the text file </li>\n<li><span translate=no>_^_1_^_</span> is the URL to download the file from </li>\n<li><span translate=no>_^_2_^_</span> is the number of characters to filter. Use this during testing when trying large datasets </li>\n<p><em>Returns</em> the text content</p></ul>\n": "<h3>\u52a0\u8f7d\u6587\u672c\u6587\u4ef6</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u6587\u672c\u6587\u4ef6\u7684\u4f4d\u7f6e</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u4ece\u4e2d\u4e0b\u8f7d\u6587\u4ef6\u7684 URL</li>\n<li><span translate=no>_^_2_^_</span>\u662f\u8981\u7b5b\u9009\u7684\u5b57\u7b26\u6570\u3002\u5728\u6d4b\u8bd5\u671f\u95f4\u5c1d\u8bd5\u5927\u578b\u6570\u636e\u96c6\u65f6\u4f7f\u7528\u6b64</li>\u9009\u9879\n<p><em>\u8fd4\u56de</em>\u6587\u672c\u5185\u5bb9</p></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Create a PyTorch tensor </p>\n": "<p>\u521b\u5efa\u4e00\u4e2a pyTorch \u5f20\u91cf</p>\n",
|
||||
"<p>Download if it doesn't exist </p>\n": "<p>\u5982\u679c\u4e0d\u5b58\u5728\uff0c\u8bf7\u4e0b\u8f7d</p>\n",
|
||||
"<p>Filter </p>\n": "<p>\u7b5b\u9009</p>\n",
|
||||
"<p>Load data </p>\n": "<p>\u52a0\u8f7d\u6570\u636e</p>\n",
|
||||
"<p>Load the content </p>\n": "<p>\u52a0\u8f7d\u5185\u5bb9</p>\n",
|
||||
"<p>Number of samples </p>\n": "<p>\u6837\u672c\u6570\u91cf</p>\n",
|
||||
"<p>Tokenize </p>\n": "<p>Tokenize</p>\n",
|
||||
"<p>Truncate </p>\n": "<p>\u622a\u65ad</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> is the list of token ids </li>\n<li><span translate=no>_^_1_^_</span> is the sequence length of a single training sample</li></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u662f\u4ee4\u724c ID \u7684\u5217\u8868</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u5355\u4e2a\u8bad\u7ec3\u6837\u672c\u7684\u5e8f\u5217\u957f\u5ea6</li></ul>\n",
|
||||
"Loads text datasets to fine-tune GPT-NeoX": "\u52a0\u8f7d\u6587\u672c\u6570\u636e\u96c6\u4ee5\u5fae\u8c03 GPT-NEOX",
|
||||
"Text Dataset for GPT-NeoX": "GPT-NEOX \u7684\u6587\u672c\u6570\u636e\u96c6"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h3>Get trainable parameters</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the model to train </li>\n<p><em>Returns</em> a list of parameters for training</p></ul>\n": "<h3>\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u53ef\u80fd\u306a\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u53d6\u5f97</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u30e2\u30c7\u30eb\u3067\u3059</li>\n<p><em>\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u7528\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u30ea\u30b9\u30c8\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Backward pass </p>\n": "<p>\u30d0\u30c3\u30af\u30ef\u30fc\u30c9\u30d1\u30b9</p>\n",
|
||||
"<p>Calculate accuracy </p>\n": "<p>\u7cbe\u5ea6\u3092\u8a08\u7b97</p>\n",
|
||||
"<p>Calculate loss </p>\n": "<p>\u640d\u5931\u306e\u8a08\u7b97</p>\n",
|
||||
"<p>Filter parameters that require gradients </p>\n": "<p>\u30b0\u30e9\u30c7\u30fc\u30b7\u30e7\u30f3\u3092\u5fc5\u8981\u3068\u3059\u308b\u30d5\u30a3\u30eb\u30bf\u30fc\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc</p>\n",
|
||||
"<p>Forward pass </p>\n": "<p>\u30d5\u30a9\u30ef\u30fc\u30c9\u30d1\u30b9</p>\n",
|
||||
"<p>Get all parameters </p>\n": "<p>\u3059\u3079\u3066\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Get predictions </p>\n": "<p>\u4e88\u6e2c\u3092\u53d6\u5f97</p>\n",
|
||||
"<p>Iterate through the batches </p>\n": "<p>\u30d0\u30c3\u30c1\u3092\u7e70\u308a\u8fd4\u3057\u51e6\u7406\u3059\u308b</p>\n",
|
||||
"<p>Move targets to the same device as output </p>\n": "<p>\u30bf\u30fc\u30b2\u30c3\u30c8\u3092\u51fa\u529b\u3068\u540c\u3058\u30c7\u30d0\u30a4\u30b9\u306b\u79fb\u52d5</p>\n",
|
||||
"<p>Optimize </p>\n": "<p>\u6700\u9069\u5316</p>\n",
|
||||
"<p>Set gradients to zero </p>\n": "<p>\u30b0\u30e9\u30c7\u30fc\u30b7\u30e7\u30f3\u3092\u30bc\u30ed\u306b\u8a2d\u5b9a</p>\n",
|
||||
"<p>Set model for train </p>\n": "<p>\u9244\u9053\u6a21\u578b\u3092\u8a2d\u5b9a</p>\n",
|
||||
"<p>tracker.add({'loss.scaled': loss}) </p>\n": "<p>\u30c8\u30e9\u30c3\u30ab\u30fc\u8ffd\u52a0 ({'\u640d\u5931.scaled': \u640d\u5931})</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> train/valid </li>\n<li><span translate=no>_^_1_^_</span> is the sample </li>\n<p><em>Returns</em> the loss, output and the target</p></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u96fb\u8eca/\u6709\u52b9</li>\n<li><span translate=no>_^_1_^_</span>\u30b5\u30f3\u30d7\u30eb\u3067\u3059</li>\n<p>\u640d\u5931\u3001\u51fa\u529b\u3001<em>\u30bf\u30fc\u30b2\u30c3\u30c8\u3092\u8fd4\u3057\u307e\u3059</em></p></ul>\n",
|
||||
"trainer.py": "trainer.py"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h3>Get trainable parameters</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the model to train </li>\n<p><em>Returns</em> a list of parameters for training</p></ul>\n": "<h3>\u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4\u0d9a\u0dc5 \u0dc4\u0dd0\u0d9a\u0dd2 \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd3\u0db1\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1</h3>\n<ul><li><span translate=no>_^_0_^_</span> \u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4 \u0d9a\u0dd2\u0dbb\u0dd3\u0db8\u0dda \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba\u0dba\u0dd2 </li>\n</ul><p>\u0db4\u0dd4\u0dc4\u0dd4\u0dab\u0dd4\u0dc0\u0dc3\u0db3\u0dc4\u0dcf \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd2 \u0dbd\u0dd0\u0dba\u0dd2\u0dc3\u0dca\u0dad\u0dd4\u0dc0\u0d9a\u0dca<em>\u0d86\u0db4\u0dc3\u0dd4</em> \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</p>\n",
|
||||
"<p> </p>\n": "<p> </p>\n",
|
||||
"<p>Backward pass </p>\n": "<p>\u0db4\u0dc3\u0dd4\u0d9c\u0dcf\u0db8\u0dd3\u0db4\u0dcf\u0dc3\u0dca </p>\n",
|
||||
"<p>Calculate accuracy </p>\n": "<p>\u0db1\u0dd2\u0dbb\u0dc0\u0daf\u0dca\u0dba\u0dad\u0dcf\u0dc0\u0dba\u0d9c\u0dab\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Calculate loss </p>\n": "<p>\u0d85\u0dbd\u0dcf\u0db7\u0dba\u0d9c\u0dab\u0db1\u0dba \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Filter parameters that require gradients </p>\n": "<p>\u0d85\u0db1\u0dd4\u0d9a\u0dca\u0dbb\u0db8\u0dd2\u0d9a\u0d85\u0dc0\u0dc1\u0dca\u0dba \u0db6\u0dc0 \u0db4\u0dd9\u0dbb\u0dc4\u0db1\u0dca \u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd3\u0db1\u0dca </p>\n",
|
||||
"<p>Forward pass </p>\n": "<p>\u0d89\u0daf\u0dd2\u0dbb\u0dd2\u0dc3\u0dcf\u0db8\u0dcf\u0dbb\u0dca\u0dae\u0dba </p>\n",
|
||||
"<p>Get all parameters </p>\n": "<p>\u0dc3\u0dd2\u0dba\u0dbd\u0dd4\u0db8\u0db4\u0dbb\u0dcf\u0db8\u0dd2\u0dad\u0dd3\u0db1\u0dca \u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Get predictions </p>\n": "<p>\u0d85\u0db1\u0dcf\u0dc0\u0dd0\u0d9a\u0dd2\u0dbd\u0db6\u0dcf \u0d9c\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Iterate through the batches </p>\n": "<p>\u0d9a\u0dab\u0dca\u0da9\u0dcf\u0dba\u0db8\u0dca\u0dc4\u0dbb\u0dc4\u0dcf \u0db1\u0dd0\u0dc0\u0dad \u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Move targets to the same device as output </p>\n": "<p>\u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba\u0dbd\u0dd9\u0dc3 \u0d91\u0d9a\u0db8 \u0d8b\u0db4\u0dcf\u0d82\u0d9c\u0dba \u0dc0\u0dd9\u0dad \u0d89\u0dbd\u0d9a\u0dca\u0d9a \u0d9c\u0dd9\u0db1\u0dba\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Optimize </p>\n": "<p>\u0db4\u0dca\u0dbb\u0dc1\u0dc3\u0dca\u0dad\u0d9a\u0dbb\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Set gradients to zero </p>\n": "<p>\u0dc1\u0dd4\u0db1\u0dca\u0dba\u0dba\u0da7\u0d85\u0db1\u0dd4\u0d9a\u0dca\u0dbb\u0db8\u0dd2\u0d9a \u0dc3\u0d9a\u0dc3\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>Set model for train </p>\n": "<p>\u0daf\u0dd4\u0db8\u0dca\u0dbb\u0dd2\u0dba\u0dc3\u0db3\u0dc4\u0dcf \u0d86\u0d9a\u0dd8\u0dad\u0dd2\u0dba \u0dc3\u0d9a\u0dc3\u0db1\u0dca\u0db1 </p>\n",
|
||||
"<p>tracker.add({'loss.scaled': loss}) </p>\n": "<p>tracker.add({'loss.scaled': \u0d85\u0dc4\u0dd2\u0db8\u0dd2 \u0dc0\u0dd3\u0db8}) </p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> train/valid </li>\n<li><span translate=no>_^_1_^_</span> is the sample </li>\n<p><em>Returns</em> the loss, output and the target</p></ul>\n": "<ul><li><span translate=no>_^_0_^_</span> \u0daf\u0dd4\u0db8\u0dca\u0dbb\u0dd2\u0dba/\u0dc0\u0dbd\u0d82\u0d9c\u0dd4 </li>\n<li><span translate=no>_^_1_^_</span> \u0db1\u0dd2\u0dba\u0dd0\u0daf\u0dd2\u0dba \u0dc0\u0dda </li>\n<p>\u0d85\u0dbd\u0dcf\u0db7\u0dba, \u0db4\u0dca\u0dbb\u0dad\u0dd2\u0daf\u0dcf\u0db1\u0dba \u0dc3\u0dc4 \u0d89\u0dbd\u0d9a\u0dca\u0d9a\u0dba<em>\u0d86\u0db4\u0dc3\u0dd4 \u0dbd\u0db6\u0dcf \u0daf\u0dd9\u0dba\u0dd2</em> </p></ul>\n",
|
||||
"trainer.py": "trainer.py"
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
{
|
||||
"<h3>Get trainable parameters</h3>\n<ul><li><span translate=no>_^_0_^_</span> is the model to train </li>\n<p><em>Returns</em> a list of parameters for training</p></ul>\n": "<h3>\u83b7\u53d6\u53ef\u8bad\u7ec3\u7684\u53c2\u6570</h3>\n<ul><li><span translate=no>_^_0_^_</span>\u662f\u8981\u8bad\u7ec3\u7684\u6a21\u578b</li>\n<p><em>\u8fd4\u56de</em>\u8bad\u7ec3\u7684\u53c2\u6570\u5217\u8868</p></ul>\n",
|
||||
"<p> </p>\n": "<p></p>\n",
|
||||
"<p>Backward pass </p>\n": "<p>\u5411\u540e\u4f20\u7403</p>\n",
|
||||
"<p>Calculate accuracy </p>\n": "<p>\u8ba1\u7b97\u7cbe\u5ea6</p>\n",
|
||||
"<p>Calculate loss </p>\n": "<p>\u8ba1\u7b97\u635f\u5931</p>\n",
|
||||
"<p>Filter parameters that require gradients </p>\n": "<p>\u8fc7\u6ee4\u9700\u8981\u6e10\u53d8\u7684\u53c2\u6570</p>\n",
|
||||
"<p>Forward pass </p>\n": "<p>\u5411\u524d\u4f20\u7403</p>\n",
|
||||
"<p>Get all parameters </p>\n": "<p>\u83b7\u53d6\u6240\u6709\u53c2\u6570</p>\n",
|
||||
"<p>Get predictions </p>\n": "<p>\u83b7\u53d6\u9884\u6d4b</p>\n",
|
||||
"<p>Iterate through the batches </p>\n": "<p>\u904d\u5386\u6279\u6b21</p>\n",
|
||||
"<p>Move targets to the same device as output </p>\n": "<p>\u5c06\u76ee\u6807\u79fb\u52a8\u5230\u4e0e\u8f93\u51fa\u76f8\u540c\u7684\u8bbe\u5907\u4e0a</p>\n",
|
||||
"<p>Optimize </p>\n": "<p>\u4f18\u5316</p>\n",
|
||||
"<p>Set gradients to zero </p>\n": "<p>\u5c06\u6e10\u53d8\u8bbe\u7f6e\u4e3a\u96f6</p>\n",
|
||||
"<p>Set model for train </p>\n": "<p>\u8bbe\u7f6e\u706b\u8f66\u6a21\u578b</p>\n",
|
||||
"<p>tracker.add({'loss.scaled': loss}) </p>\n": "<p>tracker.add ({'loss.scaled': loss})</p>\n",
|
||||
"<ul><li><span translate=no>_^_0_^_</span> train/valid </li>\n<li><span translate=no>_^_1_^_</span> is the sample </li>\n<p><em>Returns</em> the loss, output and the target</p></ul>\n": "<ul><li><span translate=no>_^_0_^_</span>\u8bad\u7ec3/\u6709\u6548</li>\n<li><span translate=no>_^_1_^_</span>\u662f\u6837\u672c</li>\n<p><em>\u8fd4\u56de</em>\u635f\u5931\u3001\u8f93\u51fa\u548c\u76ee\u6807</p></ul>\n",
|
||||
"trainer.py": "trainer.py"
|
||||
}
|
||||
Reference in New Issue
Block a user