playground/run_multinode.sh

#!/bin/bash
set -ex

CONDA_TORCHRUN=/mlx/users/jiashuo.fan/miniconda3/envs/abbie/bin/torchrun

export DISABLE_ZB=1
export USE_DETERMINISTIC_BACKWARD=0
export TORCH_NCCL_ENABLE_MONITORING=0
export WANDB_MODE=disabled
export WANDB_DISABLED=true
export UCX_ERROR_SIGNALS=""
export NCCL_IB_DISABLE=0
export UCX_TLS=rc,sm,self

cd /opt/tiger/Abbie

DATA_PATHS='[["/mnt/hdfs/byte_tt_data_cu_vagcp/haogeng.liu/new_policy7w_v2_reformat","100%"]]'

MODEL_PATH=/home/tiger/.cache/TiViLa-Qwen3-VL-8B-Instruct-Compliance
PROJNAME=qwen3_vl_cpt
EXPNAME=new_policy7w_v2_reformat
OUTDIR=/mnt/bn/bohanzhainas1/jiashuo/exp/new_policy7w_v2_reformat

LOG_DIR=/mlx/users/jiashuo.fan/playground/.claude/run_multinode_$(date +%Y%m%d_%H%M%S)
mkdir -p "$LOG_DIR"
echo "Node rank: ${ARNOLD_ID}, Nodes: ${ARNOLD_WORKER_NUM}, Master: ${METIS_WORKER_0_HOST}:${METIS_WORKER_0_PORT}"
echo "Logs: $LOG_DIR"

$CONDA_TORCHRUN \
    --nproc_per_node=${ARNOLD_WORKER_GPU} \
    --nnodes=${ARNOLD_WORKER_NUM} \
    --node_rank=${ARNOLD_ID} \
    --master_addr=${METIS_WORKER_0_HOST} \
    --master_port=${METIS_WORKER_0_PORT} \
    --log-dir="$LOG_DIR" \
tivila_trainer.py \
    model.pretrained_path=${MODEL_PATH} \
    optim.lr="1e-5" \
    optim.visual_lr="1e-5" \
    optim.lr_warmup_steps_ratio=0.03 \
    trainer.output_path=${OUTDIR} \
    trainer.project_name=${PROJNAME} \
    trainer.experiment_name=${EXPNAME} \
    trainer.log_interval=10 \
    trainer.checkpoint_interval=850 \
    trainer.checkpoint_hf_model=true \
    data.patterns=${DATA_PATHS} \
    data.transform_extra_kwargs.image_max_pixels=16777216 \
    data.transform_extra_kwargs.video_max_pixels=307200 \
    data.transform_extra_kwargs.video_max_frames=32 \
    data.max_seq_len=8192 \
    data.is_continuous_batch=true \
    data.num_training_steps=99999999 \
    data.chunks_per_step=4 \
    model.recompute_attn=true \
    model.activation_offloading=true \
    model.visual_activation_offloading=true