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

.

This commit is contained in:
hailin 2025-08-25 18:39:29 +08:00
parent 3a45712277
commit b817640e8c
1 changed files with 0 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ds_config_zero3.json
View File

@ -15,8 +15,6 @@
"reduce_bucket_size": 50000000, "reduce_bucket_size": 50000000,
"stage3_prefetch_bucket_size": 50000000, "stage3_prefetch_bucket_size": 50000000,
"stage3_param_persistence_threshold": 100000, "stage3_param_persistence_threshold": 100000,
/* 关键：保存最终权重时做 16bit 聚合，方便直接 from_pretrained 加载 */
"stage3_gather_16bit_weights_on_model_save": true "stage3_gather_16bit_weights_on_model_save": true
}, },