adapter_name_or_path: saves\LLaMA3.1-8B-Chat\lora\Luminia-8B-RP
bf16: true
cutoff_len: 4000
dataset: qa-unc-dpo
dataset_dir: data
ddp_timeout: 180000000
do_train: true
finetuning_type: lora
flash_attn: fa2
gradient_accumulation_steps: 1
include_num_input_tokens_seen: true
learning_rate: 5.0e-05
logging_steps: 10
lora_alpha: 64
lora_dropout: 0.35
lora_rank: 32
lora_target: all
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_samples: 100000
model_name_or_path: NousResearch/Meta-Llama-3.1-8B-Instruct
neftune_noise_alpha: 5
num_train_epochs: 1.0
optim: adamw_8bit
output_dir: saves\LLaMA3.1-8B-Chat\lora\Luminia-8B-RP-DPO
packing: true
per_device_train_batch_size: 1
plot_loss: true
pref_beta: 0.1
pref_ftx: 0
pref_loss: orpo
preprocessing_num_workers: 16
quantization_bit: 4
quantization_method: bitsandbytes
report_to: none
rope_scaling: linear
save_steps: 1000
stage: dpo
template: alpaca
warmup_steps: 0