ros/LlamaFactory
1
0
Fork 0
mirror of https://github.com/hiyouga/LlamaFactory.git synced 2026-03-22 12:23:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

[deps] adapt to transformers v5 (#10147)

Browse Source 
Co-authored-by: frozenleaves <frozen@Mac.local>
Co-authored-by: hiyouga <hiyouga@buaa.edu.cn>
This commit is contained in:
浮梦
2026-02-02 12:07:19 +08:00
committed by GitHub
parent 762b480131
commit bf04ca6af8
23 changed files with 149 additions and 120 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -65,7 +65,9 @@ class DataArguments:
)
mix_strategy: Literal["concat", "interleave_under", "interleave_over", "interleave_once"] = field(
default="concat",
metadata={"help": "Strategy to use in dataset mixing (concat/interleave) (undersampling/oversampling/sampling w.o. replacement)."},
metadata={
"help": "Strategy to use in dataset mixing (concat/interleave) (undersampling/oversampling/sampling w.o. replacement)."
},
)
interleave_probs: str | None = field(
default=None,

3
src/llamafactory/hparams/model_args.py
View File

@@ -206,9 +206,6 @@ class BaseModelArguments:
if self.model_name_or_path is None:
raise ValueError("Please provide `model_name_or_path`.")
if self.split_special_tokens and self.use_fast_tokenizer:
raise ValueError("`split_special_tokens` is only supported for slow tokenizers.")
if self.adapter_name_or_path is not None: # support merging multiple lora weights
self.adapter_name_or_path = [path.strip() for path in self.adapter_name_or_path.split(",")]

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

@@ -139,10 +139,6 @@ def _verify_model_args(
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.")
if data_args.template == "yi" and model_args.use_fast_tokenizer:
logger.warning_rank0("We should use slow tokenizer for the Yi models. Change `use_fast_tokenizer` to False.")
model_args.use_fast_tokenizer = False
def _check_extra_dependencies(
model_args: "ModelArguments",
@@ -188,9 +184,7 @@ def _check_extra_dependencies(
if training_args is not None:
if training_args.deepspeed:
# pin deepspeed version < 0.17 because of https://github.com/deepspeedai/DeepSpeed/issues/7347
check_version("deepspeed", mandatory=True)
check_version("deepspeed>=0.10.0,<=0.16.9")
if training_args.predict_with_generate:
check_version("jieba", mandatory=True)