Merge branch 'main' into release/2.3

Author: Jintao-Huang

docs/source/LLM/支持的模型和数据集.md

@@ -373,9 +373,12 @@
 |llava1_6-vicuna-7b-instruct|[swift/llava-v1.6-vicuna-7b-hf](https://modelscope.cn/models/swift/llava-v1.6-vicuna-7b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-vicuna|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-v1.6-vicuna-7b-hf](https://huggingface.co/llava-hf/llava-v1.6-vicuna-7b-hf)|
 |llava1_6-vicuna-13b-instruct|[swift/llava-v1.6-vicuna-13b-hf](https://modelscope.cn/models/swift/llava-v1.6-vicuna-13b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-vicuna|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-v1.6-vicuna-13b-hf](https://huggingface.co/llava-hf/llava-v1.6-vicuna-13b-hf)|
 |llava1_6-yi-34b-instruct|[swift/llava-v1.6-34b-hf](https://modelscope.cn/models/swift/llava-v1.6-34b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-yi|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-v1.6-34b-hf](https://huggingface.co/llava-hf/llava-v1.6-34b-hf)|
-|llama3-llava-next-8b|[AI-Modelscope/llama3-llava-next-8b](https://modelscope.cn/models/AI-Modelscope/llama3-llava-next-8b/summary)|^(model.layers\|model.mm_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-llava-next|✔|✘|✔|✘||vision|[lmms-lab/llama3-llava-next-8b](https://huggingface.co/lmms-lab/llama3-llava-next-8b)|
-|llava-next-72b|[AI-Modelscope/llava-next-72b](https://modelscope.cn/models/AI-Modelscope/llava-next-72b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen-instruct|✔|✘|✔|✘||vision|[lmms-lab/llava-next-72b](https://huggingface.co/lmms-lab/llava-next-72b)|
-|llava-next-110b|[AI-Modelscope/llava-next-110b](https://modelscope.cn/models/AI-Modelscope/llava-next-110b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen-instruct|✔|✘|✔|✘||vision|[lmms-lab/llava-next-110b](https://huggingface.co/lmms-lab/llava-next-110b)|
+|llama3-llava-next-8b-hf|[swift/llama3-llava-next-8b-hf](https://modelscope.cn/models/swift/llama3-llava-next-8b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-llava-next-hf|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llama3-llava-next-8b-hf](https://huggingface.co/llava-hf/llama3-llava-next-8b-hf)|
+|llava-next-72b-hf|[AI-ModelScope/llava-next-72b-hf](https://modelscope.cn/models/AI-ModelScope/llava-next-72b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-qwen-hf|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-next-72b-hf](https://huggingface.co/llava-hf/llava-next-72b-hf)|
+|llava-next-110b-hf|[AI-ModelScope/llava-next-110b-hf](https://modelscope.cn/models/AI-ModelScope/llava-next-110b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-qwen-hf|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-next-110b-hf](https://huggingface.co/llava-hf/llava-next-110b-hf)|
+|llama3-llava-next-8b|[AI-Modelscope/llama3-llava-next-8b](https://modelscope.cn/models/AI-Modelscope/llama3-llava-next-8b/summary)|^(model.layers\|model.mm_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama3-llava-next|✔|✘|✔|✘||vision|[lmms-lab/llama3-llava-next-8b](https://huggingface.co/lmms-lab/llama3-llava-next-8b)|
+|llava-next-72b|[AI-Modelscope/llava-next-72b](https://modelscope.cn/models/AI-Modelscope/llava-next-72b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen|✔|✘|✔|✘||vision|[lmms-lab/llava-next-72b](https://huggingface.co/lmms-lab/llava-next-72b)|
+|llava-next-110b|[AI-Modelscope/llava-next-110b](https://modelscope.cn/models/AI-Modelscope/llava-next-110b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen|✔|✘|✔|✘||vision|[lmms-lab/llava-next-110b](https://huggingface.co/lmms-lab/llava-next-110b)|
 |llava-next-video-7b-instruct|[swift/LLaVA-NeXT-Video-7B-hf](https://modelscope.cn/models/swift/LLaVA-NeXT-Video-7B-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-next-video|✔|✘|✘|✘|transformers>=4.42, av|video|[llava-hf/LLaVA-NeXT-Video-7B-hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-hf)|
 |llava-next-video-7b-32k-instruct|[swift/LLaVA-NeXT-Video-7B-32K-hf](https://modelscope.cn/models/swift/LLaVA-NeXT-Video-7B-32K-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-next-video|✔|✘|✘|✘|transformers>=4.42, av|video|[llava-hf/LLaVA-NeXT-Video-7B-32K-hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-32K-hf)|
 |llava-next-video-7b-dpo-instruct|[swift/LLaVA-NeXT-Video-7B-DPO-hf](https://modelscope.cn/models/swift/LLaVA-NeXT-Video-7B-DPO-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-next-video|✔|✘|✘|✘|transformers>=4.42, av|video|[llava-hf/LLaVA-NeXT-Video-7B-DPO-hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-DPO-hf)|

docs/source_en/LLM/Supported-models-datasets.md

@@ -373,9 +373,12 @@ The table below introcudes all models supported by SWIFT:
 |llava1_6-vicuna-7b-instruct|[swift/llava-v1.6-vicuna-7b-hf](https://modelscope.cn/models/swift/llava-v1.6-vicuna-7b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-vicuna|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-v1.6-vicuna-7b-hf](https://huggingface.co/llava-hf/llava-v1.6-vicuna-7b-hf)|
 |llava1_6-vicuna-13b-instruct|[swift/llava-v1.6-vicuna-13b-hf](https://modelscope.cn/models/swift/llava-v1.6-vicuna-13b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-vicuna|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-v1.6-vicuna-13b-hf](https://huggingface.co/llava-hf/llava-v1.6-vicuna-13b-hf)|
 |llava1_6-yi-34b-instruct|[swift/llava-v1.6-34b-hf](https://modelscope.cn/models/swift/llava-v1.6-34b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-yi|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-v1.6-34b-hf](https://huggingface.co/llava-hf/llava-v1.6-34b-hf)|
-|llama3-llava-next-8b|[AI-Modelscope/llama3-llava-next-8b](https://modelscope.cn/models/AI-Modelscope/llama3-llava-next-8b/summary)|^(model.layers\|model.mm_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-llava-next|✔|✘|✔|✘||vision|[lmms-lab/llama3-llava-next-8b](https://huggingface.co/lmms-lab/llama3-llava-next-8b)|
-|llava-next-72b|[AI-Modelscope/llava-next-72b](https://modelscope.cn/models/AI-Modelscope/llava-next-72b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen-instruct|✔|✘|✔|✘||vision|[lmms-lab/llava-next-72b](https://huggingface.co/lmms-lab/llava-next-72b)|
-|llava-next-110b|[AI-Modelscope/llava-next-110b](https://modelscope.cn/models/AI-Modelscope/llava-next-110b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen-instruct|✔|✘|✔|✘||vision|[lmms-lab/llava-next-110b](https://huggingface.co/lmms-lab/llava-next-110b)|
+|llama3-llava-next-8b-hf|[swift/llama3-llava-next-8b-hf](https://modelscope.cn/models/swift/llama3-llava-next-8b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-llava-next-hf|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llama3-llava-next-8b-hf](https://huggingface.co/llava-hf/llama3-llava-next-8b-hf)|
+|llava-next-72b-hf|[AI-ModelScope/llava-next-72b-hf](https://modelscope.cn/models/AI-ModelScope/llava-next-72b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-qwen-hf|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-next-72b-hf](https://huggingface.co/llava-hf/llava-next-72b-hf)|
+|llava-next-110b-hf|[AI-ModelScope/llava-next-110b-hf](https://modelscope.cn/models/AI-ModelScope/llava-next-110b-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama-qwen-hf|✔|✔|✘|✘|transformers>=4.39|vision|[llava-hf/llava-next-110b-hf](https://huggingface.co/llava-hf/llava-next-110b-hf)|
+|llama3-llava-next-8b|[AI-Modelscope/llama3-llava-next-8b](https://modelscope.cn/models/AI-Modelscope/llama3-llava-next-8b/summary)|^(model.layers\|model.mm_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llama3-llava-next|✔|✘|✔|✘||vision|[lmms-lab/llama3-llava-next-8b](https://huggingface.co/lmms-lab/llama3-llava-next-8b)|
+|llava-next-72b|[AI-Modelscope/llava-next-72b](https://modelscope.cn/models/AI-Modelscope/llava-next-72b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen|✔|✘|✔|✘||vision|[lmms-lab/llava-next-72b](https://huggingface.co/lmms-lab/llava-next-72b)|
+|llava-next-110b|[AI-Modelscope/llava-next-110b](https://modelscope.cn/models/AI-Modelscope/llava-next-110b/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-qwen|✔|✘|✔|✘||vision|[lmms-lab/llava-next-110b](https://huggingface.co/lmms-lab/llava-next-110b)|
 |llava-next-video-7b-instruct|[swift/LLaVA-NeXT-Video-7B-hf](https://modelscope.cn/models/swift/LLaVA-NeXT-Video-7B-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-next-video|✔|✘|✘|✘|transformers>=4.42, av|video|[llava-hf/LLaVA-NeXT-Video-7B-hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-hf)|
 |llava-next-video-7b-32k-instruct|[swift/LLaVA-NeXT-Video-7B-32K-hf](https://modelscope.cn/models/swift/LLaVA-NeXT-Video-7B-32K-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-next-video|✔|✘|✘|✘|transformers>=4.42, av|video|[llava-hf/LLaVA-NeXT-Video-7B-32K-hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-32K-hf)|
 |llava-next-video-7b-dpo-instruct|[swift/LLaVA-NeXT-Video-7B-DPO-hf](https://modelscope.cn/models/swift/LLaVA-NeXT-Video-7B-DPO-hf/summary)|^(language_model\|multi_modal_projector)(?!.\*(lm_head\|output\|emb\|wte\|shared)).\*|llava-next-video|✔|✘|✘|✘|transformers>=4.42, av|video|[llava-hf/LLaVA-NeXT-Video-7B-DPO-hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-DPO-hf)|

swift/cli/main.py

@@ -53,19 +53,11 @@ def cli_main() -> None:
         argv = ['--rlhf_type', method_name] + argv
     file_path = importlib.util.find_spec(ROUTE_MAPPING[method_name]).origin
     torchrun_args = get_torchrun_args()
+    python_cmd = sys.executable
     if torchrun_args is None or method_name not in ('pt', 'sft', 'dpo', 'orpo', 'simpo', 'rlhf'):
-        try:
-            python_cmd = 'python'
-            subprocess.run(
-                [python_cmd, '--version'],
-                stdout=subprocess.PIPE,
-                stderr=subprocess.PIPE,
-            )
-        except FileNotFoundError:
-            python_cmd = 'python3'
         args = [python_cmd, file_path, *argv]
     else:
-        args = ['torchrun', *torchrun_args, file_path, *argv]
+        args = [python_cmd, '-m', 'torch.distributed.run', *torchrun_args, file_path, *argv]
     print(f"run sh: `{' '.join(args)}`", flush=True)
     result = subprocess.run(args)
     if result.returncode != 0:

swift/llm/rlhf.py

@@ -191,6 +191,7 @@ def llm_rlhf(args: RLHFArguments) -> Dict[str, Any]:
     You can also use the --model_type parameter to specify the  template.')
         template: Template = get_template(
             'chatml', tokenizer, args.system, args.max_length, args.truncation_strategy, model=model)
+    template._is_training = True
     args.system = template.default_system
     logger.info(f'system: {args.system}')
 

swift/llm/sft.py

@@ -296,6 +296,7 @@ def llm_sft(args: SftArguments) -> Dict[str, Any]:
         args.truncation_strategy,
         model=model,
         **template_kwargs)
+    template._is_training = True
     if streaming:
         template.encode = partial(template.encode, streaming=streaming)
     args.system = template.default_system

swift/llm/utils/argument.py

@@ -988,6 +988,10 @@ def __post_init__(self) -> None:
                 self.learning_rate = 1e-5
             if self.save_only_model is None:
                 self.save_only_model = True
+                logger.warning(
+                    'Due to the adoption of full-parameter training, '
+                    'in order to avoid saving excessive weights, we set save_only_model to True. '
+                    'If you want to resume training from a checkpoint, please manually pass `--save_only_model false`.')
             if self.eval_steps is None:
                 self.eval_steps = 200
         else: