pythia410m-sft-tldr / code /configs /dpo1b_relabel_generated_same_prompts.yml
mnoukhov's picture
Training in progress, step 500
1904ee8 verified
raw
history blame contribute delete
360 Bytes
output_dir: /home/toolkit/huggingface/openai_comparisons_20k_regen_and_relabelled
mode: relabel
model_name: mnoukhov/pythia410m-tldrprompt-dpo1b-adapter
dataset_name: arianhosseini/openai_comparisons_20k_regen_and_relabelled
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