LLaMA-Factory/examples/extras/badam/train_single_gpu.sh
Jonery bc1c082bc2 add example
Former-commit-id: 97c523516093961983037922e7fc84e4010d5fec
2024-06-18 13:50:26 +08:00

37 lines
985 B
Bash

#!/bin/bash
export CUDA_VISIBLE_DEVICES=0
cd ../../..
llamafactory-cli train \
--stage sft \
--do_train True \
--model_name_or_path meta-llama/Llama-2-13b-hf \
--preprocessing_num_workers 16 \
--finetuning_type full \
--template default \
--flash_attn auto \
--dataset_dir data \
--dataset alpaca_en_demo \
--cutoff_len 1024 \
--learning_rate 1e-6 \
--num_train_epochs 3.0 \
--max_samples 100000 \
--per_device_train_batch_size 1 \
--gradient_accumulation_steps 8 \
--lr_scheduler_type cosine \
--max_grad_norm 1.0 \
--logging_steps 5 \
--save_steps 100 \
--warmup_steps 0 \
--optim adamw_torch \
--packing False \
--report_to none \
--use_badam True \
--output_dir saves/LLaMA2-13B/full/BAdam \
--plot_loss True \
--ddp_timeout 180000000 \
--include_num_input_tokens_seen True \
--badam_mode layer \
--badam_switch_mode ascending \
--badam_switch_interval 50