chore: import upstream snapshot with attribution

tests/models/test_flash_attn.py

@@ -0,0 +1,8 @@
+from swift.model import get_model_processor
+
+if __name__ == '__main__':
+    # model, tokenizer = get_model_processor('Qwen/Qwen2-7B-Instruct', attn_impl='flash_attn')
+    # model, tokenizer = get_model_processor('AIDC-AI/Ovis2-2B', attn_impl='flash_attn')
+    # model, tokenizer = get_model_processor('OpenGVLab/InternVL2-2B', attn_impl='flash_attn')
+    model, tokenizer = get_model_processor('Shanghai_AI_Laboratory/internlm3-8b-instruct', attn_impl='flash_attn')
+    print(model)

tests/models/test_llm.py

@@ -0,0 +1,17 @@
+import os
+
+os.environ['CUDA_VISIBLE_DEVICES'] = '0,1'
+os.environ['ASCEND_RT_VISIBLE_DEVICES'] = '0,1'
+
+
+def test_llama3():
+    from swift import InferArguments, infer_main
+    infer_main(
+        InferArguments(
+            model='LLM-Research/Meta-Llama-3.1-8B-Instruct',
+            max_batch_size=2,
+            val_dataset='AI-ModelScope/alpaca-gpt4-data-en#2'))
+
+
+if __name__ == '__main__':
+    test_llama3()

tests/models/test_mllm.py

@@ -0,0 +1,19 @@
+import os
+
+os.environ['CUDA_VISIBLE_DEVICES'] = '0'
+os.environ['ASCEND_RT_VISIBLE_DEVICES'] = '0'
+
+
+def test_cogvlm():
+    from swift import InferArguments, SftArguments, infer_main, sft_main
+
+    # infer_main(InferArguments(model='ZhipuAI/cogvlm2-video-llama3-chat'))
+    sft_main(
+        SftArguments(
+            model='ZhipuAI/cogvlm2-video-llama3-chat',
+            dataset=['AI-ModelScope/alpaca-gpt4-data-zh#200', 'swift/VideoChatGPT:Generic#200'],
+            split_dataset_ratio=0.01))
+
+
+if __name__ == '__main__':
+    test_cogvlm()