bf16: true
cutoff_len: 2048
dataset: coder4boptim
dataset_dir: /root/LLaMA-Factory/data
ddp_timeout: 180000000
do_train: true
double_quantization: true
enable_thinking: true
finetuning_type: lora
flash_attn: auto
freeze_multi_modal_projector: true
freeze_vision_tower: true
gradient_accumulation_steps: 8
image_max_pixels: 589824
image_min_pixels: 1024
include_num_input_tokens_seen: true
learning_rate: 5.0e-05
logging_steps: 5
lora_alpha: 16
lora_dropout: 0.01
lora_rank: 8
lora_target: all
loraplus_lr_ratio: 16
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_samples: 6000
model_name_or_path: /root/autodl-tmp/Qwen3.5-4B-Thinking
num_train_epochs: 1.0
optim: adamw_torch
output_dir: saves/Qwen3.5-4B-Thinking/lora/aiAIL_coding_V1
packing: false
per_device_train_batch_size: 2
plot_loss: true
preprocessing_num_workers: 16
quantization_bit: 4
quantization_method: bnb
report_to: none
save_steps: 100
stage: sft
template: qwen3_5
trust_remote_code: true
video_max_pixels: 65536
video_min_pixels: 256
warmup_steps: 0