fix #3602

examples/full_multi_gpu/multi_node.sh

@@ -1,6 +1,12 @@
 #!/bin/bash
 
-python -m torch.distributed.run \
+NPROC_PER_NODE=4
+NNODES=2
+RANK=0
+MASTER_ADDR=192.168.0.1
+MASTER_PORT=29500
+
+CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.run \
     --nproc_per_node $NPROC_PER_NODE \
     --nnodes $NNODES \
     --node_rank $RANK \

examples/full_multi_gpu/single_node.sh

@@ -1,4 +1,9 @@
 #!/bin/bash
 
-deepspeed --include "localhost:0,1,2,3" \
+NPROC_PER_NODE=4
+
+CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.run \
+    --nproc_per_node $NPROC_PER_NODE \
+    --nnodes 1 \
+    --standalone \
     src/train.py examples/full_multi_gpu/llama3_full_sft.yaml