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

.

This commit is contained in:
hailin 2025-08-28 22:54:46 +08:00
parent 4ad2be2d34
commit 1e989784d3
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
train_sft_ds.py
View File

@ -487,7 +487,7 @@ def main():
torch_dtype=dtype, torch_dtype=dtype,
low_cpu_mem_usage=True, low_cpu_mem_usage=True,
trust_remote_code=True, trust_remote_code=True,
attn_implementation="flash_attention_2" attn_implementation="sdpa"
) )
print(f"GC enabled? {getattr(model, 'is_gradient_checkpointing', False)}", flush=True) print(f"GC enabled? {getattr(model, 'is_gradient_checkpointing', False)}", flush=True)