mirror of
https://github.com/hanzoai/open-instruct.git
synced 2026-10-30 18:26:32 +00:00
Yizhong WangandGitHub
2dcfd43a3a
* Bump transformers and flash-attn versions to support buil-in flash attention in Transformers. * Update demo finetuning scripts. * Fix model names and saving dirs in training scripts. * Update hf_trainer finetuning code, supporting flash attention.
46 lines
1.5 KiB
Bash
Executable File
46 lines
1.5 KiB
Bash
Executable File
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
|
|
|
|
MODEL_SIZE=70B
|
|
NUM_GPUS=8
|
|
BATCH_SIZE_PER_GPU=1
|
|
TOTAL_BATCH_SIZE=128
|
|
GRADIENT_ACC_STEPS=$(($TOTAL_BATCH_SIZE/$NUM_GPUS/$BATCH_SIZE_PER_GPU))
|
|
echo "Training llama model ${MODEL_SIZE} using $NUM_GPUS GPUs, $BATCH_SIZE_PER_GPU batch size per GPU, $GRADIENT_ACC_STEPS gradient accumulation steps"
|
|
|
|
# Lora training
|
|
accelerate launch \
|
|
--num_machines 1 \
|
|
--num_processes $NUM_GPUS \
|
|
open_instruct/finetune.py \
|
|
--model_name_or_path ../hf_llama2_models/${MODEL_SIZE} \
|
|
--gradient_checkpointing \
|
|
--use_qlora \
|
|
--use_lora \
|
|
--use_flash_attn \
|
|
--lora_rank 64 \
|
|
--lora_alpha 16 \
|
|
--lora_dropout 0.1 \
|
|
--tokenizer_name ../hf_llama2_models/${MODEL_SIZE} \
|
|
--use_slow_tokenizer \
|
|
--train_file data/processed/tulu_v2/tulu_v2_data.jsonl \
|
|
--max_seq_length 4096 \
|
|
--preprocessing_num_workers 128 \
|
|
--per_device_train_batch_size $BATCH_SIZE_PER_GPU \
|
|
--gradient_accumulation_steps $GRADIENT_ACC_STEPS \
|
|
--learning_rate 1e-4 \
|
|
--lr_scheduler_type linear \
|
|
--warmup_ratio 0.03 \
|
|
--weight_decay 0. \
|
|
--num_train_epochs 5 \
|
|
--output_dir output/tulu_v2_${MODEL_SIZE}_qlora/ \
|
|
--with_tracking \
|
|
--report_to tensorboard \
|
|
--logging_steps 1 &&
|
|
|
|
python open_instruct/merge_lora.py \
|
|
--base_model_name_or_path ../hf_llama2_models/${MODEL_SIZE} \
|
|
--lora_model_name_or_path output/tulu_v2_${MODEL_SIZE}_qlora/ \
|
|
--output_dir output/tulu_v2_${MODEL_SIZE}_qlora_merged/ \
|
|
--qlora \
|
|
--save_tokenizer
|