### model
model_name_or_path: meta-llama/Llama-3.1-8B-Instruct
reward_model: chchen/Llama-3.1-8B-Instruct-reward-500
trust_remote_code: true

### method
stage: ppo
do_train: true
finetuning_type: lora
lora_target: all

### dataset
dataset: bct_non_cot_sft_500
dataset_dir: data_private
template: llama3
cutoff_len: 1024
# max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16

### output
output_dir: saves/Llama-3.1-8B-Instruct/ppo-500/train
logging_steps: 10
save_steps: 50
plot_loss: true
overwrite_output_dir: true
save_total_limit: 3
push_to_hub: true
hub_model_id: chchen/Llama-3.1-8B-Instruct-2410-ppo-500

### train
per_device_train_batch_size: 4
gradient_accumulation_steps: 8
learning_rate: 1.0e-5
num_train_epochs: 10.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000

### generate
max_new_tokens: 512
top_k: 0
top_p: 0.9