mirror of
https://github.com/SakanaAI/doc-to-lora.git
synced 2026-04-25 00:06:20 +02:00
30 lines
991 B
Bash
Executable file
30 lines
991 B
Bash
Executable file
#!/bin/bash
|
|
|
|
port=29051
|
|
|
|
uv run accelerate launch --config_file accelerate_config.yaml --main_process_port $port \
|
|
--num_processes=8 --gpu_ids all train.py \
|
|
configs/main_exp/self_gen_lv1_closed_qa_1_l2l.yaml \
|
|
--model_name_or_path=google/gemma-2-2b-it \
|
|
--target_modules=down_proj \
|
|
--lora_r=8 \
|
|
--eval_strategy=no \
|
|
--max_qas_len=512 \
|
|
--max_qas_per_sample=1 \
|
|
--per_rank_gen=True \
|
|
--per_layer_processing=True \
|
|
--gen_lora_l1_reg_coef=0.1 \
|
|
--max_steps=20000 \
|
|
--gradient_accumulation_steps=16 \
|
|
--max_packed_inp_len=1024 \
|
|
--max_packed_ctx_len=2048 \
|
|
--use_per_ctx_average_loss=True \
|
|
--use_kl_loss=True \
|
|
--quantize_ctx_encoder=True \
|
|
--torch_empty_cache_steps=10 \
|
|
--from_pretrained_checkpoint=train_outputs/runs/$RUN_NAME/checkpoint-80000/pytorch_model.bin \
|
|
--max_ctx_chunk_len=512 \
|
|
--min_ctx_chunk_len=25 \
|
|
--num_chunk_probs='{"1":"0.5", "2":"0.125", "3":"0.0625", "4":"0.0625", "5":"0.0625", "6":"0.0625", "7":"0.0625", "8":"0.0625"}' \
|
|
--warmup_steps=2000 \
|
|
--learning_rate=2e-5
|