diff --git a/scripts/train_sft.sh b/scripts/train_sft.sh index 6f5e97d..2d331dc 100644 --- a/scripts/train_sft.sh +++ b/scripts/train_sft.sh @@ -22,5 +22,7 @@ export PYTHONUNBUFFERED=1 # 禁止日志缓存(CLAUDE.md §5 export PYTORCH_ALLOC_CONF=expandable_segments:True # 变长序列 batch 易碎片化,按需扩段 export HF_ENDPOINT=${HF_ENDPOINT:-https://hf-mirror.com} export HF_HOME=${HF_HOME:-/data/zym/hf_cache} # 模型缓存落 /data,根分区已满 +# hf-mirror 不代理 HF Xet CAS(大权重走 Xet 会 401,见 train_whitebox.sh 详注) +export HF_HUB_DISABLE_XET=1 torchrun --nproc_per_node=4 --master_port=29571 scripts/train_sft.py "$MODE" diff --git a/scripts/train_whitebox.sh b/scripts/train_whitebox.sh index 6b3f528..ee46507 100755 --- a/scripts/train_whitebox.sh +++ b/scripts/train_whitebox.sh @@ -27,5 +27,9 @@ export PYTHONUNBUFFERED=1 # 禁止日志缓存(CLAUDE.md §5 export PYTORCH_ALLOC_CONF=expandable_segments:True # 变长生成序列易碎片化,按需扩段 export HF_ENDPOINT=${HF_ENDPOINT:-https://hf-mirror.com} export HF_HOME=${HF_HOME:-/data/zym/hf_cache} # 模型缓存落 /data,根分区已满 +# 非显然坑:hf-mirror 不代理 HF 的 Xet CAS——大权重走 Xet 会直连 +# cas-server.xethub.hf.co 并返 401(2026-07-19 teacher 4B 下载实撞)。禁用 Xet +# 退回经典 HTTP/LFS 下载(镜像支持)。若仍不行:pip uninstall hf_xet +export HF_HUB_DISABLE_XET=1 torchrun --nproc_per_node=4 --master_port=29572 scripts/train_whitebox.py "$MODE"