text2sql_step_1_lora_sft.yaml

### model
model_name_or_path: 
  /home/data2/luzhan/projects/local_models/deepseek-ai/deepseek-coder-6.7b-instruct
quantization_bit: 4

### method
stage: sft
do_train: true
finetuning_type: lora
lora_target: q_proj,v_proj

### dataset
dataset: text2sql_step_1
template: deepseekcoder
cutoff_len: 4096
max_samples: 100000
overwrite_cache: true
preprocessing_num_workers: 16

### output
output_dir: /home/data2/luzhan/projects/LLM-Empowered-Text2SQL/finetuned_model
logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true

### train
per_device_train_batch_size: 1
gradient_accumulation_steps: 8
learning_rate: 0.00001
num_train_epochs: 3.0
lr_scheduler_type: cosine
warmup_steps: 0.1
fp16: true

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