423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-27 08:12:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix #1073 #1462 #1735 #1908

Browse Source 
Former-commit-id: 31165a9822bd52130b33cd3439f887c26e0679dc
This commit is contained in:
hiyouga 2023-12-20 17:15:40 +08:00
parent f64be8ee84
commit 82a79e9fdf
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/llmtuner/train/sft/workflow.py
View File

@ -32,6 +32,9 @@ def run_sft(
if training_args.predict_with_generate:
tokenizer.padding_side = "left" # use left-padding in generation
if getattr(model, "is_quantized", False) and not training_args.do_train:
setattr(model, "_hf_peft_config_loaded", True) # hack here: make model compatible with prediction
data_collator = DataCollatorForSeq2Seq(
tokenizer=tokenizer,
pad_to_multiple_of=8 if tokenizer.padding_side == "right" else None, # for shift short attention