423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-11-04 18:02:19 +08:00
Code Issues Packages Projects Releases Wiki Activity

add Yi-VL-34B model

Browse Source 
Former-commit-id: 8b3d8a7e3bd8dff27cc72edba1b8a042f6d1929c
This commit is contained in:
hiyouga 2024-05-15 22:58:19 +08:00
parent 4d7f0fbb7a
commit 2ac972d6e7
5 changed files with 15 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@ -166,7 +166,7 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
| [StarCoder2](https://huggingface.co/bigcode) | 3B/7B/15B | q_proj,v_proj | - |
| [XVERSE](https://huggingface.co/xverse) | 7B/13B/65B | q_proj,v_proj | xverse |
| [Yi (1/1.5)](https://huggingface.co/01-ai) | 6B/9B/34B | q_proj,v_proj | yi |
| [Yi-VL](https://huggingface.co/01-ai) | 6B | q_proj,v_proj | yi_vl |
| [Yi-VL](https://huggingface.co/01-ai) | 6B/34B | q_proj,v_proj | yi_vl |
| [Yuan](https://huggingface.co/IEITYuan) | 2B/51B/102B | q_proj,v_proj | yuan |
> [!NOTE]

2
README_zh.md
View File

@ -166,7 +166,7 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
| [StarCoder2](https://huggingface.co/bigcode) | 3B/7B/15B | q_proj,v_proj | - |
| [XVERSE](https://huggingface.co/xverse) | 7B/13B/65B | q_proj,v_proj | xverse |
| [Yi (1/1.5)](https://huggingface.co/01-ai) | 6B/9B/34B | q_proj,v_proj | yi |
| [Yi-VL](https://huggingface.co/01-ai) | 6B | q_proj,v_proj | yi_vl |
| [Yi-VL](https://huggingface.co/01-ai) | 6B/34B | q_proj,v_proj | yi_vl |
| [Yuan](https://huggingface.co/IEITYuan) | 2B/51B/102B | q_proj,v_proj | yuan |
> [!NOTE]

3
src/llmtuner/extras/constants.py
View File

@ -1215,6 +1215,9 @@ register_model_group(
"YiVL-6B-Chat": {
DownloadSource.DEFAULT: "BUAADreamer/Yi-VL-6B-hf",
},
"YiVL-34B-Chat": {
DownloadSource.DEFAULT: "BUAADreamer/Yi-VL-34B-hf",
},
},
template="yi_vl",
vision=True,

11
src/llmtuner/model/loader.py
View File

@ -78,8 +78,15 @@ def load_tokenizer(model_args: "ModelArguments") -> "TokenizerModule":
patch_tokenizer(tokenizer)
if model_args.visual_inputs:
processor = AutoProcessor.from_pretrained(model_args.model_name_or_path, **init_kwargs)
setattr(processor, "tokenizer", tokenizer)
try:
processor = AutoProcessor.from_pretrained(model_args.model_name_or_path, **init_kwargs)
setattr(processor, "tokenizer", tokenizer)
except Exception:
raise ValueError(
"This multimodal LLM is not supported.\n"
"Download LLaVA-1.5 models from: https://huggingface.co/llava-hf\n"
"Download Yi-VL models from: https://huggingface.co/BUAADreamer"
)
else:
processor = None

2
src/llmtuner/model/utils/visual.py
View File

@ -58,7 +58,7 @@ class LlavaMultiModalProjectorForYiVLForVLLM(LlavaMultiModalProjectorForYiVL):
self.linear_2 = torch.nn.LayerNorm(text_hidden_size, bias=True)
self.linear_3 = torch.nn.Linear(text_hidden_size, text_hidden_size, bias=True)
self.linear_4 = torch.nn.LayerNorm(text_hidden_size, bias=True)
self.act = torch.nn.GELU()
self.act = ACT2FN[projector_hidden_act]
def autocast_projector_dtype(