compute_accuracy: 'true'
cutoff_len: 1024
dataset: JB_Detect_v2
dataset_dir: data
ddp_timeout: 180000000
do_sample: 'false'
do_train: true
eval_steps: 5000
eval_strategy: steps
finetuning_type: lora
flash_attn: auto
fp16: true
gradient_accumulation_steps: 8
learning_rate: 5.0e-05
logging_steps: 100
lora_alpha: 128
lora_dropout: 0.05
lora_rank: 64
lora_target: all
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_new_tokens: 2
max_samples: 10000000000
model_name_or_path: Qwen/Qwen2.5-0.5B-Instruct
num_train_epochs: 1.0
optim: adamw_torch
output_dir: saves/Qwen2.5-0.5B-Instruct/lora/train_2025-05-29-20-20-04_2
packing: false
per_device_eval_batch_size: 8
per_device_train_batch_size: 8
plot_loss: true
preprocessing_num_workers: 16
report_to: none
save_steps: 5000
stage: sft
template: qwen
val_size: 0.08
warmup_steps: 1239