14 lines
568 B
YAML
14 lines
568 B
YAML
|
|
model_name: mnoukhov/dpo2_costa_1b_20k_fp16.yml_dff3275532270a8cbadb56d184c5d31d
|
||
|
|
wandb_log_id: model_name
|
||
|
|
base_model_name: vwxyzjn/EleutherAI_pythia-1b-deduped__sft__tldr
|
||
|
|
base_model_revision: sft__55513__1706646024
|
||
|
|
gen_dtype: bfloat16
|
||
|
|
tokenizer_name: EleutherAI/pythia-1b-deduped
|
||
|
|
dataset_name: vwxyzjn/summarize_from_feedback_tldr_3_filtered_oai_preprocessing_1706381144
|
||
|
|
gold_model_name: vwxyzjn/EleutherAI_pythia-6.9b-deduped__reward__tldr
|
||
|
|
gold_model_revision: reward__55513__1708628552
|
||
|
|
eval_dtype: bfloat16
|
||
|
|
max_new_tokens: 128
|
||
|
|
max_length: 640
|
||
|
|
temperature: 0.010001
|