diff --git a/README.md b/README.md index cc1e6ba..6bf1d44 100644 --- a/README.md +++ b/README.md @@ -1,2 +1,5 @@ -# hh_rlhf +--- +{} +--- +# hh_rlhf \ No newline at end of file diff --git a/config.json b/config.json new file mode 100644 index 0000000..e97829a --- /dev/null +++ b/config.json @@ -0,0 +1,56 @@ +{ + "model": null, + "signature": { + "description": "Reason step-by-step to determine which response is better for the given context.\nAs you reason, note any uncertainties, ambiguities, and gaps \u2014 both in how the instructions apply to the task at hand and in whether you have the knowledge needed to solve it. Be honest in your reasoning when you are unsure about your answer. Your reasoning must lead directly to\nthe better response (A or B).", + "properties": { + "question": { + "__dspy_field_type": "input", + "desc": "The question or prompt provided.", + "prefix": "Question:", + "title": "Question", + "type": "string" + }, + "response_A": { + "__dspy_field_type": "input", + "desc": "The first response option.", + "prefix": "Response A:", + "title": "Response A", + "type": "string" + }, + "response_B": { + "__dspy_field_type": "input", + "desc": "The second response option.", + "prefix": "Response B:", + "title": "Response B", + "type": "string" + }, + "reasoning": { + "__dspy_field_type": "output", + "desc": "Step-by-step reasoning, including uncertainty, ambiguity, or uncertainty in your thought process when relevant.", + "prefix": "Reasoning:", + "title": "Reasoning", + "type": "string" + }, + "label": { + "__dspy_field_type": "output", + "desc": "The label of the better response ('A' or 'B').", + "enum": [ + "A", + "B" + ], + "prefix": "Label:", + "title": "Label", + "type": "string" + } + }, + "required": [ + "question", + "response_A", + "response_B", + "reasoning", + "label" + ], + "title": "PreferenceJudgeSig", + "type": "object" + } +} \ No newline at end of file diff --git a/program.json b/program.json new file mode 100644 index 0000000..1962771 --- /dev/null +++ b/program.json @@ -0,0 +1,48 @@ +{ + "traces": [], + "train": [], + "demos": [], + "signature": { + "instructions": "Reason step-by-step to determine which response is better for the given context.\nAs you reason, note any uncertainties, ambiguities, and gaps \u2014 both in how the instructions apply to the task at hand and in whether you have the knowledge needed to solve it. Be honest in your reasoning when you are unsure about your answer. Your reasoning must lead directly to\nthe better response (A or B).", + "fields": [ + { + "prefix": "Question:", + "description": "The question or prompt provided." + }, + { + "prefix": "Response A:", + "description": "The first response option." + }, + { + "prefix": "Response B:", + "description": "The second response option." + }, + { + "prefix": "Reasoning:", + "description": "Step-by-step reasoning, including uncertainty, ambiguity, or uncertainty in your thought process when relevant." + }, + { + "prefix": "Label:", + "description": "The label of the better response ('A' or 'B')." + } + ] + }, + "lm": { + "model": "together_ai/Qwen/Qwen2.5-7B-Instruct-Turbo", + "model_type": "chat", + "cache": true, + "num_retries": 3, + "finetuning_model": null, + "launch_kwargs": {}, + "train_kwargs": {}, + "temperature": null, + "max_tokens": null + }, + "metadata": { + "dependency_versions": { + "python": "3.11", + "dspy": "3.1.3", + "cloudpickle": "3.1" + } + } +} \ No newline at end of file