From d8e8c2e2c88cb81b18cfeacf5d0b77f25b9b80e3 Mon Sep 17 00:00:00 2001 From: Tyrin Todd Date: Thu, 23 Apr 2026 23:05:27 -0700 Subject: [PATCH] server-inference push ultrafeedback --- README.md | 5 ++++- config.json | 51 +++++++++++++++++++++++++++++++++++++++++++++++++++ program.json | 44 ++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 99 insertions(+), 1 deletion(-) create mode 100644 config.json create mode 100644 program.json diff --git a/README.md b/README.md index 0ec2bcd..a18eaf0 100644 --- a/README.md +++ b/README.md @@ -1,2 +1,5 @@ -# ultrafeedback +--- +{} +--- +# ultrafeedback \ No newline at end of file diff --git a/config.json b/config.json new file mode 100644 index 0000000..9ca9729 --- /dev/null +++ b/config.json @@ -0,0 +1,51 @@ +{ + "model": null, + "signature": { + "description": "Rate only instruction-following quality for the candidate response on a 1-5 scale.\nUse 1 for very poor instruction following and 5 for excellent instruction following. As 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.", + "properties": { + "prompt": { + "__dspy_field_type": "input", + "desc": "The user instruction/prompt.", + "prefix": "Prompt:", + "title": "Prompt", + "type": "string" + }, + "response": { + "__dspy_field_type": "input", + "desc": "The model response to rate.", + "prefix": "Response:", + "title": "Response", + "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" + }, + "instruction_following": { + "__dspy_field_type": "output", + "desc": "Instruction-following rating on a 1-5 scale.", + "enum": [ + "1", + "2", + "3", + "4", + "5" + ], + "prefix": "Instruction Following:", + "title": "Instruction Following", + "type": "string" + } + }, + "required": [ + "prompt", + "response", + "reasoning", + "instruction_following" + ], + "title": "UltraFeedbackInstructionJudgeSig", + "type": "object" + } +} \ No newline at end of file diff --git a/program.json b/program.json new file mode 100644 index 0000000..750a5b5 --- /dev/null +++ b/program.json @@ -0,0 +1,44 @@ +{ + "traces": [], + "train": [], + "demos": [], + "signature": { + "instructions": "Rate only instruction-following quality for the candidate response on a 1-5 scale.\nUse 1 for very poor instruction following and 5 for excellent instruction following. As 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.", + "fields": [ + { + "prefix": "Prompt:", + "description": "The user instruction/prompt." + }, + { + "prefix": "Response:", + "description": "The model response to rate." + }, + { + "prefix": "Reasoning:", + "description": "Step-by-step reasoning, including uncertainty, ambiguity, or uncertainty in your thought process when relevant." + }, + { + "prefix": "Instruction Following:", + "description": "Instruction-following rating on a 1-5 scale." + } + ] + }, + "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