lora_sft.yaml

### model
model_name_or_path: /root/autodl-tmp/LLM-Research/Phi-3-medium-128k-instruct

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

### dataset
dataset: 
template: phi
cutoff_len: 1024
max_samples: 100000
overwrite_cache: true
preprocessing_num_workers: 64

### output
output_dir: saves/Phi-3-medium-128k-instruct/lora/sft
logging_steps: 10
save_steps: 3000
plot_loss: true
overwrite_output_dir: true

### train
per_device_train_batch_size: 3
gradient_accumulation_steps: 8
learning_rate: 1.0e-4
num_train_epochs: 3.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000

### eval
# val_size: 0.1
# per_device_eval_batch_size: 1
# eval_strategy: steps
# eval_steps: 500