mirror of
https://github.com/hiyouga/LLaMA-Factory.git
synced 2025-08-02 03:32:50 +08:00
7 lines
222 B
Bash
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
|