hailin
/
jd_train
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

.

This commit is contained in:
hailin 2025-08-26 14:31:55 +08:00
parent fa4d4d8b1e
commit c063c2dc06
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
train_sft_ds.py
View File

@ -315,6 +315,7 @@ def parse_args():
ap.add_argument("--local_rank", type=int, default=-1, ap.add_argument("--local_rank", type=int, default=-1,
help="for deepspeed/torchrun launcher; ignored by user code") help="for deepspeed/torchrun launcher; ignored by user code")
ap.add_argument("--per_device_eval_batch_size", type=int, default=1) ap.add_argument("--per_device_eval_batch_size", type=int, default=1)
ap.add_argument("--deepspeed", type=str, default=None)
return ap.parse_args() return ap.parse_args()
@ -590,7 +591,7 @@ def main():
logging_steps=args.log_interval, logging_steps=args.log_interval,
save_steps=args.save_steps, save_steps=args.save_steps,
save_total_limit=2, save_total_limit=2,
deepspeed=args.deepspeed, deepspeed=(args.deepspeed if args.deepspeed and os.path.isfile(args.deepspeed) else None),
dataloader_drop_last=False, # 关键：别丢尾，避免空 batch dataloader_drop_last=False, # 关键：别丢尾，避免空 batch
dataloader_num_workers=0, dataloader_num_workers=0,
dataloader_prefetch_factor=None, dataloader_prefetch_factor=None,