run_robotwin_train.sh

# export NCCL_SOCKET_IFNAME=bond0
# export NCCL_IB_HCA=mlx5_2,mlx5_3

# used for check save when communication
export NCCL_BLOCKING_WAIT=1
export NCCL_ASYNC_ERROR_HANDLING=1
export NCCL_TIMEOUT=1000  # timeout set to 1 hour (unit: seconds)
export NCCL_SOCKET_TIMEOUT_MS=360000
export NCCL_P2P_DISABLE=1
# export NCCL_DEBUG=INFO
# export NCCL_DEBUG_SUBSYS=ALL
# export TORCH_DISTRIBUTED_DEBUG=DETAIL

###########################################################################################
# === Please modify the following paths according to your environment ===
Framework_name=QwenOFT
freeze_module_list=''
base_vlm=/inspire/ssd/project/embodied-basic-model/zhangjianing-253108140206/DATASET/model/spiritv1.5
config_yaml=./examples/Robotwin/train_files/starvla_cotrain_robotwin.yaml
robotwin_data_root=/inspire/ssd/project/embodied-basic-model/zhangjianing-253108140206/DATASET/robotwin_lerobot
run_root_dir=/inspire/ssd/project/embodied-basic-model/zhangjianing-253108140206/experiment/spirit_vla/starvla-vla/results
data_mix=robotwin
run_id=124_${data_mix}_spirit
# === End of environment variable configuration ===
###########################################################################################

#batchsize=24
export WANDB_MODE=disabled

output_dir=${run_root_dir}/${run_id}
mkdir -p ${output_dir}
# mv this script to the output dir
cp $0 ${output_dir}/
#这里的数据没有put_object_dustbin和scan objects 改了mixtures
#bash examples/Robotwin/train_files/run_robotwin_train.sh
accelerate launch \
  --config_file starVLA/config/deepseeds/deepspeed_zero2.yaml \
  --num_processes 4 \
  starVLA/training/train_starvla.py \
  --config_yaml ${config_yaml} \
  --framework.name ${Framework_name} \
  --framework.qwenvl.base_vlm ${base_vlm} \
  --datasets.vla_data.per_device_batch_size 8 \
  --datasets.vla_data.data_mix ${data_mix} \
  --datasets.vla_data.data_root_dir ${robotwin_data_root}\
  --trainer.freeze_modules ${freeze_module_list} \
  --trainer.max_train_steps 30000 \
  --trainer.save_interval 10000 \
  --trainer.logging_frequency 100 \
  --trainer.eval_interval 1000 \
  --run_root_dir ${run_root_dir} \
  --run_id ${run_id} \
  --wandb_project spirit \
  --wandb_entity 1732949190-tongji-university  \
  # --is_debug True



##### Multi-Server Multi-GPU training script #####
  # accelerate launch \
  #   --config_file starVLA/config/deepseeds/deepspeed_zero2.yaml \
  #   --main_process_ip $MASTER_ADDR \
  #   --main_process_port $MASTER_PORT \
  #   --machine_rank $SLURM_PROCID \
  #   --num_machines $SLURM_NNODES \
  #   --num_processes=${TOTAL_GPUS} \
  #   starVLA/training/train_starvla.py \
  #   --config_yaml ${config_yaml} \
  #   --framework.name ${Framework_name} \
  #   --framework.qwenvl.base_vlm ${base_vlm} \
  #   --run_root_dir ${run_root_dir} \
  #   --run_id ${run_id} \
  #   --wandb_project your_project \
  #   --wandb_entity your_name
##### Multi-Server Multi-GPU training script #####