cutoff_len: 2048
dataset: docker_command_dataset
dataset_dir: data
ddp_timeout: 180000000
do_train: true
double_quantization: true
finetuning_type: lora
flash_attn: auto
fp16: true
gradient_accumulation_steps: 8
include_num_input_tokens_seen: true
learning_rate: 0.0002
logging_steps: 5
lora_alpha: 16
lora_dropout: 0
lora_rank: 8
lora_target: all
lr_scheduler_type: cosine
max_grad_norm: 0.3
max_samples: 100000
model_name_or_path: llava-hf/llava-v1.6-mistral-7b-hf
num_train_epochs: 3.0
optim: adamw_torch
output_dir: saves/LLaVA-NeXT-Mistral-7B-Chat/lora/train_2025-04-21-17-35-28
packing: false
per_device_train_batch_size: 2
plot_loss: true
preprocessing_num_workers: 16
quantization_bit: 4
quantization_method: bitsandbytes
report_to: none
save_steps: 100
stage: sft
template: llava_next_mistral
trust_remote_code: true
warmup_steps: 0