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

[misc] fix packing and eval plot (#7623)

Browse Source
This commit is contained in:
hoshi-hiyouga
2025-04-07 18:20:57 +08:00
committed by GitHub
parent 5115dc8c7f
commit c3c0efbaa0
70 changed files with 288 additions and 194 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llamafactory/hparams/parser.py
View File

@@ -135,7 +135,7 @@ def _check_extra_dependencies(
check_version("mixture-of-depth>=1.1.6", mandatory=True)
if model_args.infer_backend == EngineName.VLLM:
check_version("vllm>=0.4.3,<=0.8.2")
check_version("vllm>=0.4.3,<=0.8.3")
check_version("vllm", mandatory=True)
elif model_args.infer_backend == EngineName.SGLANG:
check_version("sglang>=0.4.4")
@@ -285,10 +285,6 @@ def get_train_args(args: Optional[Union[dict[str, Any], list[str]]] = None) -> _
if model_args.use_unsloth and is_deepspeed_zero3_enabled():
raise ValueError("Unsloth is incompatible with DeepSpeed ZeRO-3.")
if data_args.neat_packing and not data_args.packing:
logger.warning_rank0("`neat_packing` requires `packing` is True. Change `packing` to True.")
data_args.packing = True
_verify_model_args(model_args, data_args, finetuning_args)
_check_extra_dependencies(model_args, finetuning_args, training_args)