wehub/labmlai--annotated_deep_learning_paper_implementations
1
0
Fork 0
Code Issues Pull Requests Actions Deployments Models Agent Notes Packages Projects Releases Wiki Activity

chore: import upstream snapshot with attribution

Browse Source
This commit is contained in:
wehub-resource-sync
2026-07-13 12:19:01 +08:00
commit 3b90d1192f
2172 changed files with 594509 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
translate_cache/neox/evaluation/__init__.ja.json
+54
View File
@@ -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&#x27;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"
}
translate_cache/neox/evaluation/__init__.si.json
+54
View File
@@ -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&#x27;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"
}
translate_cache/neox/evaluation/__init__.zh.json
+54
View File
@@ -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&#x27;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"
}
translate_cache/neox/evaluation/half_precision.ja.json
+10
View File
@@ -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"
}
translate_cache/neox/evaluation/half_precision.si.json
+10
View File
@@ -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"
}
translate_cache/neox/evaluation/half_precision.zh.json
+10
View File
@@ -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"
}
translate_cache/neox/evaluation/llm_int8.ja.json
+11
View File
@@ -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"
}
translate_cache/neox/evaluation/llm_int8.si.json
+11
View File
@@ -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"
}
translate_cache/neox/evaluation/llm_int8.zh.json
+11
View File
@@ -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"
}