2.2.0

ascend-robotfix(pytorch):Ensure no PP/VPP stage contains only empty layers during LoRA fine-tuning

337813b0创建于 4月2日历史提交

文件	最后提交记录	最后更新时间
tokenizer	!2880 [pytorch][feature]add qwen3 reasoning template	11 个月前
__init__.py	!2945 [pytorch][refactor] refactor the patch framework	10 个月前
arguments.py	[pytorch][bugfix] vpp and layer-list is conflict	6 个月前
checkpointing.py	[pytorch][bugfix] update checkpoint loading for LoRA and QLoRA	7 个月前
initialize.py	!3154 [pytorch][feature]the parameter --use-mc2 has been renamed to --use-ascend-mc2	9 个月前
one_logger_utils.py	!3195 [pytorch][feature]add elastic-training feature	9 个月前
training.py	[pytorch][bugfix]fix about cleancode	8 个月前
utils.py	fix(pytorch):Ensure no PP/VPP stage contains only empty layers during LoRA fine-tuning	2 个月前