423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-29 10:10:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

[model] update kt code (#9406)

Browse Source
This commit is contained in:
Yaowei Zheng
2025-11-05 15:27:22 +08:00
committed by GitHub
parent 56f45e826f
commit eaf963f67f
28 changed files with 108 additions and 68 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llamafactory/train/ksft/workflow.py
View File

@@ -47,6 +47,7 @@ def run_sft(
model = load_model(tokenizer, model_args, finetuning_args, training_args.do_train)
from ktransformers.util.globals import GLOBAL_CONFIG
GLOBAL_CONFIG._config["mod"] = "sft"
if getattr(model, "is_quantized", False) and not training_args.do_train:
@@ -66,12 +67,13 @@ def run_sft(
# Metric utils
metric_module = {}
if training_args.predict_with_generate:
raise NotImplementedError("`predict_with_generate` is not supported in KTransformers SFT yet. if you do need it, please open an issue.")
raise NotImplementedError("`predict_with_generate` is not supported in KTransformers SFT yet.")
elif finetuning_args.compute_accuracy:
raise NotImplementedError("`compute_accuracy` is not supported in KTransformers SFT yet. if you do need it, please open an issue.")
raise NotImplementedError("`compute_accuracy` is not supported in KTransformers SFT yet.")
# Initialize our Trainer
from ktransformers.sft.lora import KTrainer
trainer = KTrainer(
model=model,
args=training_args,