cutoff_len: 1024
dataset: identity
dataset_dir: data
do_train: true
finetuning_type: lora
flash_attn: auto
fp16: true
gradient_accumulation_steps: 8
learning_rate: 5.0e-05
logging_steps: 5
lora_alpha: 16
lora_dropout: 0
lora_rank: 8
lora_target: q_proj,v_proj
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_samples: 100000
model_name_or_path: Qwen/Qwen1.5-0.5B-Chat
num_train_epochs: 3.0
optim: adamw_torch
output_dir: saves/Qwen1.5-0.5B-Chat/lora/QwenTT-0.5B-INT8
packing: false
per_device_train_batch_size: 2
plot_loss: true
preprocessing_num_workers: 16
quantization_bit: 8
report_to: none
save_steps: 100
stage: sft
template: qwen
warmup_steps: 0