LLaMA-Factory/examples/extras/fsdp_qlora/single_node.sh

7 lines
222 B
Bash

#!/bin/bash
# DO NOT use GPTQ/AWQ model in FSDP+QLoRA
CUDA_VISIBLE_DEVICES=0,1 accelerate launch \
--config_file examples/accelerate/fsdp_config.yaml \
src/train.py examples/extras/fsdp_qlora/llama3_lora_sft.yaml