output_dir: openai_summarize_vllm_generated_20k_label410m
mode: relabel
model_name: mnoukhov/pythia410m-tldrprompt-dpo1b-adapter
dataset_name: mnoukhov/openai_summarize_vllm_generated_20k
eval_split: train
use_peft: False
beta: 0.5
load_in_8bit: False
bf16: False
fp16: True
per_device_eval_batch_size: 8
warmup_steps: 150