423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-30 02:30:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

[model] add qwen2.5 vl models (#6779)

Browse Source 
Former-commit-id: 999c7c8fe0
This commit is contained in:
hoshi-hiyouga
2025-01-31 03:00:29 +08:00
committed by GitHub
parent f6779b0e0c
commit e8c1979b79
8 changed files with 77 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/llamafactory/chat/hf_engine.py
View File

@@ -176,7 +176,10 @@ class HuggingfaceEngine(BaseEngine):
if torch.is_floating_point(value): # cast data dtype for paligemma
value = value.to(model.dtype)
gen_kwargs[key] = value.to(model.device)
if key == "second_per_grid_ts": # qwen2.5vl special case
gen_kwargs[key] = value.tolist()
else:
gen_kwargs[key] = value.to(model.device)
if getattr(model.config, "model_type", None) in ["minicpmv", "minicpmo"]:
gen_kwargs["input_ids"] = inputs