423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2026-05-28 10:58:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

[refactor] Add KTransformers AMX MoE SFT support via Accelerate (#10430)

Browse Source 
Co-authored-by: mrhaoxx <mr.haoxx@gmail.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Peilin Li
2026-05-01 01:47:58 +08:00
committed by GitHub
parent 6b08b948c9
commit 887ee2b121
39 changed files with 287 additions and 1968 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/llamafactory/hparams/parser.py
View File

@@ -186,13 +186,16 @@ def _verify_model_args(
raise ValueError("Quantized model only accepts a single adapter. Merge them first.")
def _check_extra_dependencies(
model_args: "ModelArguments",
finetuning_args: "FinetuningArguments",
training_args: Optional["TrainingArguments"] = None,
) -> None:
if model_args.use_kt:
check_version("ktransformers", mandatory=True)
check_version("kt-kernel", mandatory=True)
check_version("transformers-kt", mandatory=True)
check_version("accelerate-kt", mandatory=True)
if model_args.use_unsloth:
check_version("unsloth", mandatory=True)
@@ -510,6 +513,9 @@ def get_train_args(args: dict[str, Any] | list[str] | None = None) -> _TRAIN_CLS
)
transformers.set_seed(training_args.seed)
if model_args.use_kt:
model_args.apply_kt_config(finetuning_args, training_args, model_args.model_max_length)
return model_args, data_args, training_args, finetuning_args, generating_args