Support InternLM3 Dense 8B Model (#6640)

* support internlm3

* update

* update

* update

* add hint

Former-commit-id: deacc00b1226ca3d53bf7bb1231cf276eaa8296b

README.md

@@ -87,6 +87,7 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
 </details>
 
 ## Changelog
+[25/01/15] We supported fine-tuning the **[InternLM3](https://huggingface.co/collections/internlm/)** models. Thank [@hhaAndroid](https://github.com/hhaAndroid)'s PR.
 
 [25/01/10] We supported fine-tuning the **[Phi-4](https://huggingface.co/microsoft/phi-4)** model.
 
@@ -201,6 +202,7 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
 | [Granite 3.0-3.1](https://huggingface.co/ibm-granite)             | 1B/2B/3B/8B                      | granite3         |
 | [Index](https://huggingface.co/IndexTeam)                         | 1.9B                             | index            |
 | [InternLM2/InternLM2.5](https://huggingface.co/internlm)          | 7B/20B                           | intern2          |
+| [InternLM3](https://huggingface.co/internlm)                      | 8B                               | intern3          |
 | [Llama](https://github.com/facebookresearch/llama)                | 7B/13B/33B/65B                   | -                |
 | [Llama 2](https://huggingface.co/meta-llama)                      | 7B/13B/70B                       | llama2           |
 | [Llama 3-3.3](https://huggingface.co/meta-llama)                  | 1B/3B/8B/70B                     | llama3           |

README_zh.md

@@ -89,6 +89,8 @@ https://github.com/user-attachments/assets/e6ce34b0-52d5-4f3e-a830-592106c4c272
 
 ## 更新日志
 
+[25/01/15] 我们支持了 **[InternLM3](https://huggingface.co/collections/internlm/)** 模型的微调。感谢 [@hhaAndroid](https://github.com/hhaAndroid) 的 PR。
+
 [25/01/10] 我们支持了 **[Phi-4](https://huggingface.co/microsoft/phi-4)** 模型的微调。
 
 [24/12/21] 我们支持了使用 **[SwanLab](https://github.com/SwanHubX/SwanLab)** 跟踪与可视化实验。详细用法请参考 [此部分](#使用-swanlab-面板)。
@@ -202,6 +204,7 @@ https://github.com/user-attachments/assets/e6ce34b0-52d5-4f3e-a830-592106c4c272
 | [Granite 3.0-3.1](https://huggingface.co/ibm-granite)             | 1B/2B/3B/8B                      | granite3         |
 | [Index](https://huggingface.co/IndexTeam)                         | 1.9B                             | index            |
 | [InternLM2/InternLM2.5](https://huggingface.co/internlm)          | 7B/20B                           | intern2          |
+| [InternLM3](https://huggingface.co/internlm)                      | 8B                               | intern3          |
 | [Llama](https://github.com/facebookresearch/llama)                | 7B/13B/33B/65B                   | -                |
 | [Llama 2](https://huggingface.co/meta-llama)                      | 7B/13B/70B                       | llama2           |
 | [Llama 3-3.3](https://huggingface.co/meta-llama)                  | 1B/3B/8B/70B                     | llama3           |

src/llamafactory/data/template.py

@@ -732,6 +732,14 @@ _register_template(
     stop_words=["<|im_end|>"],
 )
 
+_register_template(
+    name="intern3",
+    format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_assistant=StringFormatter(slots=["{{content}}<|im_end|>\n"]),
+    format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    stop_words=["<|im_end|>"]
+)
 
 _register_template(
     name="llama2",

src/llamafactory/extras/constants.py

@@ -817,6 +817,15 @@ register_model_group(
     template="intern2",
 )
 
+register_model_group(
+    models={
+        "InternLM3-8B-Chat": {
+            DownloadSource.DEFAULT: "internlm/internlm3-8b-instruct",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm3-8b-instruct",
+        },
+    },
+    template="intern3",
+)
 
 register_model_group(
     models={

src/llamafactory/model/patcher.py

@@ -24,6 +24,7 @@ from transformers.modeling_utils import is_fsdp_enabled
 
 from ..extras import logging
 from ..extras.misc import infer_optim_dtype
+from ..extras.packages import is_transformers_version_greater_than
 from .model_utils.attention import configure_attn_implementation, print_attn_implementation
 from .model_utils.checkpointing import prepare_model_for_training
 from .model_utils.embedding import resize_embedding_layer
@@ -117,6 +118,9 @@ def patch_config(
     if "LlavaLlamaForCausalLM" in getattr(config, "architectures", []):
         raise ValueError("Please download llava models with hf-compatible format: https://huggingface.co/llava-hf")
 
+    if getattr(config, "model_type", None) == "internlm3" and not is_transformers_version_greater_than("4.47.1"):
+        raise RuntimeError("InternLM3 model requires transformers >= 4.47.1, please upgrade it.")
+
     # deepspeed zero3 is not compatible with low_cpu_mem_usage
     init_kwargs["low_cpu_mem_usage"] = model_args.low_cpu_mem_usage and (not is_deepspeed_zero3_enabled())