| exec >> /workspace/_polish_train.log 2>&1 | |
| echo "[$(date)] TRAIN WRAPPER START" | |
| source /opt/miniconda3/etc/profile.d/conda.sh | |
| conda activate dp | |
| cd /workspace/diffusion-pipe | |
| export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True | |
| export NCCL_P2P_DISABLE=1 | |
| export NCCL_IB_DISABLE=1 | |
| export NCCL_DEBUG=WARN | |
| export NCCL_TIMEOUT=600 | |
| export NCCL_BLOCKING_WAIT=1 | |
| export NCCL_ASYNC_ERROR_HANDLING=1 | |
| exec setsid deepspeed --num_gpus=8 train.py --deepspeed --config /workspace/configs/38_anima_polish.toml | |