start_torch_fp8.sh

#!/bin/bash

export NCCL_ALGO=Ring
export NCCL_PROTO=Simple
export MP=8
export CONFIG=config.json

torchrun --nproc-per-node ${MP} generate.py --ckpt-path /path/of/DeepSeek-V4-Flash-FP8-MP8 --config ${CONFIG} --interactive