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

.

This commit is contained in:
hailin 2025-09-14 14:57:33 +08:00
parent 80a763fa69
commit d809ada414
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
merge_zero3_safetensors.sh
View File

@ -3,7 +3,7 @@ set -euo pipefail
# ===== 可调参数 ===== # ===== 可调参数 =====
CKPT_ROOT="/home/test/checkpoints/q3-32b-lora" # 若实际是 .../checkpoint-62/global_step62，请把 CKPT_ROOT 改成 .../checkpoint-62 CKPT_ROOT="/home/test/checkpoints/q3-32b-lora" # 若实际是 .../checkpoint-62/global_step62，请把 CKPT_ROOT 改成 .../checkpoint-62
TAG="global_step20" TAG="global_step30"
HOSTS=(tn01 tn02 tn03 tn04 tn05 tn06) HOSTS=(tn01 tn02 tn03 tn04 tn05 tn06)
AGGREGATOR_HOST="tn06" # 本脚本运行/汇总所在机器 AGGREGATOR_HOST="tn06" # 本脚本运行/汇总所在机器
EXPECTED_SHARDS_PER_HOST=4 # 每机应写出分片数（按你的并行布局） EXPECTED_SHARDS_PER_HOST=4 # 每机应写出分片数（按你的并行布局）