Update train.sh
Browse files
train.sh
CHANGED
|
@@ -46,6 +46,7 @@ nohup torchrun --nproc_per_node=8 --nnodes=2 --node_rank=$NODE_RANK --master_add
|
|
| 46 |
ssh root@30.207.98.112 "
|
| 47 |
source ~/.bashrc; \
|
| 48 |
conda activate yuan; \
|
|
|
|
| 49 |
|
| 50 |
# 设置与主节点相同的NCCL环境变量
|
| 51 |
export NCCL_IB_GID_INDEX=3
|
|
|
|
| 46 |
ssh root@30.207.98.112 "
|
| 47 |
source ~/.bashrc; \
|
| 48 |
conda activate yuan; \
|
| 49 |
+
cd $TRAIN_DIR; \
|
| 50 |
|
| 51 |
# 设置与主节点相同的NCCL环境变量
|
| 52 |
export NCCL_IB_GID_INDEX=3
|