operators:
- id: load_in_code
  config:
    type: load_preexisting
    framework_name: load_in_code_0d
- id: sample_dataset_final
  config:
    type: function
    function: data_strategies.commons.uniform_sample_fixed
    function_config:
      num_samples: 3_950
  input_ids:
    - load_in_code
- id: add_row_ids
  config:
    type: function
    function: data_strategies.commons.add_row_ids
    function_config:
      column_name: ms_id
  input_ids:
    - sample_dataset_final
- id: duplicate
  config:
    type: function
    function: data_strategies.TestTime.generator.duplicate_rows
    function_config:
      n_copies: 16
  input_ids:
    - add_row_ids
- id: annotate_r1
  config:
    type: completions
    map: deepseek_reasoner
    map_config:
      input_problem_column: instruction_seed
    model: deepseek-reasoner
    batch: false
    temperature: 0.999
    require_all_responses: false
    backend: openai
    backend_params:
      max_retries: 3
      max_requests_per_minute: 400
      max_tokens_per_minute: 1_000_000_000
      base_url: "https://api.deepseek.com/"
      invalid_finish_reasons: ['content_filter']
  input_ids:
    - duplicate
- id: convert_reasoning_trace_to_final
  config:
    type: function
    function: data_strategies.commons.convert_reasoning_trace_to_final
    function_config:
      reasoning_column: reasoning
      solution_column: deepseek_solution
      output_column: final_reasoning_trace
  input_ids:
  - annotate_r1
- id: convert_to_sharegpt
  config:
    type: function
    function: data_strategies.commons.convert_instruction_response_to_sharegpt
    function_config:
      input_instruction_column: instruction_seed
      input_response_column: final_reasoning_trace
      output_sharegpt_column: conversations
  input_ids:
    - convert_reasoning_trace_to_final