source examples/fsdp2/env_config.sh

NPUS_PER_NODE=8
MASTER_ADDR=localhost
MASTER_PORT=12322
NNODES=1
NODE_RANK=0
WORLD_SIZE=$(($NPUS_PER_NODE*$NNODES))

DISTRIBUTED_ARGS="
    --nproc_per_node $NPUS_PER_NODE \
    --nnodes $NNODES \
    --node_rank $NODE_RANK \
    --master_addr $MASTER_ADDR \
    --master_port $MASTER_PORT
"

torchrun $DISTRIBUTED_ARGS train_fsdp2.py ./tests/st/shell_scripts/pretrain_qwen3_8b_4K_fsdp2.yaml