LLaMA-Factory-Mirror/examples/lora_single_gpu/llama3_preprocess.yaml

22 lines
423 B
YAML
Raw Normal View History

2024-05-17 01:02:00 +08:00
### model
2024-05-06 21:47:00 +08:00
model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
2024-05-17 01:02:00 +08:00
### method
2024-05-06 21:47:00 +08:00
stage: sft
do_train: true
finetuning_type: lora
lora_target: q_proj,v_proj
2024-05-17 01:02:00 +08:00
### dataset
2024-05-06 21:47:00 +08:00
dataset: identity,alpaca_gpt4_en
template: llama3
cutoff_len: 1024
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
2024-05-06 22:51:02 +08:00
tokenized_path: saves/llama3-8b/dataset/sft
2024-05-06 21:47:00 +08:00
2024-05-17 01:02:00 +08:00
### output
2024-05-06 21:47:00 +08:00
output_dir: saves/llama3-8b/lora/sft
overwrite_output_dir: true