423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-23 06:12:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

Bug Fix: off is parsed as False in yaml file, changed to disabled to avoid this.

Browse Source 
Former-commit-id: 3ed063f281d1c2563df1b9eb3800543208c9dc16
This commit is contained in:
stceum 2024-06-24 20:39:20 +08:00
parent bb9f48590f
commit 9aa640f27b
3 changed files with 6 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/llamafactory/hparams/model_args.py
View File

@ -97,7 +97,7 @@ class ModelArguments:
default=None, default=None,
metadata={"help": "Which scaling strategy should be adopted for the RoPE embeddings."}, metadata={"help": "Which scaling strategy should be adopted for the RoPE embeddings."},
) )
flash_attn: Literal["off", "sdpa", "fa2", "auto"] = field( flash_attn: Literal["disabled", "sdpa", "fa2", "auto"] = field(
default="auto", default="auto",
metadata={"help": "Enable FlashAttention for faster training and inference."}, metadata={"help": "Enable FlashAttention for faster training and inference."},
) )

4
src/llamafactory/hparams/parser.py
View File

@ -102,6 +102,10 @@ def _verify_model_args(model_args: "ModelArguments", finetuning_args: "Finetunin
if model_args.adapter_name_or_path is not None and len(model_args.adapter_name_or_path) != 1: if model_args.adapter_name_or_path is not None and len(model_args.adapter_name_or_path) != 1:
raise ValueError("Quantized model only accepts a single adapter. Merge them first.") raise ValueError("Quantized model only accepts a single adapter. Merge them first.")
# In case that `flash_attn` is set to `off` in the yaml file, and parsed as `False` afterwards.
if model_args.flash_attn == False:
raise ValueError("flash_attn should be \"disabled\", \"sdpa\", \"fa2\" or \"auto\".")
def _check_extra_dependencies( def _check_extra_dependencies(
model_args: "ModelArguments", model_args: "ModelArguments",

2
src/llamafactory/model/model_utils/attention.py
View File

@ -32,7 +32,7 @@ def configure_attn_implementation(config: "PretrainedConfig", model_args: "Model
if model_args.flash_attn == "auto": if model_args.flash_attn == "auto":
return return
elif model_args.flash_attn == "off": elif model_args.flash_attn == "disabled":
requested_attn_implementation = "eager" requested_attn_implementation = "eager"
elif model_args.flash_attn == "sdpa": elif model_args.flash_attn == "sdpa":