diff --git a/rewardbench/utils.py b/rewardbench/utils.py index a5f818f..69b17bd 100644 --- a/rewardbench/utils.py +++ b/rewardbench/utils.py @@ -483,7 +483,7 @@ def map_conversations_ift(example): ] return example - dataset = raw_dataset.map( + dataset = unrolled_dataset.map( map_conversations_ift, # fn_kwargs={"core_set": core_set}, num_proc=8, diff --git a/scripts/configs/eval_configs.yaml b/scripts/configs/eval_configs.yaml index cfaf42c..15f5844 100644 --- a/scripts/configs/eval_configs.yaml +++ b/scripts/configs/eval_configs.yaml @@ -615,6 +615,15 @@ Ahjeong/MMPO_Gemma_7b: batch_size: 2 trust_remote_code: False dpo: True +Ahjeong/MMPO_Gemma_7b_gamma1.1_epoch3: + model: Ahjeong/MMPO_Gemma_7b_gamma1.1_epoch3 + ref_model: kykim0/gemma-7b-ultrachat-sft + tokenizer: Ahjeong/MMPO_Gemma_7b_gamma1.1_epoch3 + chat_template: # none for tokenizer + num_gpus: 2 + batch_size: 2 + trust_remote_code: False + dpo: True wenbopan/Faro-Yi-9B-DPO: model: wenbopan/Faro-Yi-9B-DPO ref_model: wenbopan/Faro-Yi-9B diff --git a/scripts/run_generative.py b/scripts/run_generative.py index ea423a6..0786dbd 100644 --- a/scripts/run_generative.py +++ b/scripts/run_generative.py @@ -145,6 +145,8 @@ def main(): is_prometheus = True elif "gemini" in args.model: model_modifier = "gemini" + else: + model_modifier = None ############################ # Load dataset