#!/bin/bash # source .venv/bin/activate export CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES:-0} export TORCH_HOME=$TORCH_HOME # e.g. /hfm/user/torch_cache export HF_HOME=$HF_HOME # e.g. /hfm/user/torch_cache/huggingface export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True source .venv/bin/activate echo "Training with $nprocs GPUs, which is/are $CUDA_VISIBLE_DEVICES" python InternVLA/deploy/internvla_serve_sim.py \ --host 0.0.0.0 \ --port 22085 \ --checkpoint-path $CHECKPOINT_PATH # e.g. .../***.pt