cutoff_len: 6100
dataset: longzu
dataset_dir: C:\AI\LLaMA-Factory\data
ddp_timeout: 180000000
do_train: true
finetuning_type: full
flash_attn: auto
gradient_accumulation_steps: 8
include_num_input_tokens_seen: true
learning_rate: 5.0e-05
logging_steps: 5
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_samples: 100000
model_name_or_path: C:\AI\Qwen2_0.5B
num_train_epochs: 35.0
optim: adamw_torch
output_dir: saves\Qwen2-0.5B\full\train_2024-06-08-23-23-14
packing: false
per_device_train_batch_size: 1
plot_loss: true
preprocessing_num_workers: 16
report_to: none
save_steps: 100
stage: pt
template: default
warmup_steps: 0