bf16: true
cutoff_len: 4096
dataset: gpt4_judge,Judge
dataset_dir: data
ddp_timeout: 180000000
do_train: true
finetuning_type: lora
flash_attn: auto
gradient_accumulation_steps: 8
include_num_input_tokens_seen: true
learning_rate: 5.0e-05
logging_steps: 5
lora_alpha: 16
lora_dropout: 0
lora_rank: 16
lora_target: all
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_samples: 100000
model_name_or_path: /home/marl/.cache/huggingface/hub/models--meta-llama--Meta-Llama-3.1-8B-Instruct/snapshots/07eb05b21d191a58c577b4a45982fe0c049d0693
num_train_epochs: 10.0
optim: adamw_torch
output_dir: saves/LLaMA3-8B-Chat/lora/JudgePierce
packing: false
per_device_train_batch_size: 2
pissa_convert: true
pissa_init: true
plot_loss: true
preprocessing_num_workers: 16
quantization_method: bitsandbytes
report_to: none
rope_scaling: linear
save_steps: 100
stage: sft
template: llama3
use_dora: true
use_rslora: true
use_unsloth: true
warmup_steps: 0