p04798526
/
LLaMA-Factory-Mirror
1
0
Fork 1
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix patcher

This commit is contained in:
hiyouga 2024-03-15 19:18:42 +08:00
parent 113cc04719
commit 85c376fc1e
1 changed files with 1 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/llmtuner/model/patcher.py
View File

@ -279,13 +279,11 @@ def patch_config(
model_args.compute_dtype = infer_optim_dtype(model_dtype=getattr(config, "torch_dtype", None))
if getattr(config, "model_type", None) == "qwen":
setattr(config, "use_flash_attn", model_args.flash_attn)
for dtype_name, dtype in [("fp16", torch.float16), ("bf16", torch.bfloat16), ("fp32", torch.float32)]:
setattr(config, dtype_name, model_args.compute_dtype == dtype)
_configure_attn_implementation(model_args, init_kwargs)
if getattr(config, "model_type", None) == "qwen" and init_kwargs["attn_implementation"] != "flash_attention_2":
config.use_flash_attn = False
_configure_rope(config, model_args, is_trainable)
_configure_longlora(config, model_args, is_trainable)
_configure_quantization(config, tokenizer, model_args, init_kwargs)