| operators: | |
| - id: load_numina | |
| config: | |
| type: hf_source | |
| dataset: AI-MO/NuminaMath-CoT | |
| split: train | |
| - id: select_olympiad | |
| config: | |
| type: function | |
| function: data_strategies.commons.select_rows | |
| function_config: | |
| conditions: | |
| source: olympiads | |
| input_ids: | |
| - load_numina | |
| - id: sample_dataset | |
| config: | |
| type: function | |
| function: data_strategies.commons.uniform_sample_limited | |
| function_config: | |
| num_samples: 20 # 1.25% of 10000 target from stage 1 | |
| input_ids: | |
| - select_olympiad | |
| - id: get_evolution_prompt | |
| config: | |
| type: function | |
| function: data_strategies.OpenMathInstruct.utils.create_new_prompt | |
| function_config: | |
| question_column: problem | |
| input_ids: | |
| - sample_dataset | |
| - id: generate_new_questions | |
| config: | |
| type: completions | |
| model: gpt-4o-mini | |
| map: chat | |
| map_config: | |
| user_message_column: prompt_for_new_question | |
| output_column: response | |
| output_column: new_question | |
| temperature: 0.7 | |
| batch: false | |
| input_ids: | |
| - get_evolution_prompt | |