operators:
- id: load_preexisting
  config:
    type: load_preexisting
    framework_name: d1_code_load_in_r1
- id: drop_columns
  config:
    type: function
    function: data_strategies.commons.remove_columns
    function_config:
      columns:
        - conversations
  input_ids:
    - load_preexisting
- id: get_question_answer_string
  config:
    type: function
    function: data_strategies.QuestionAnswerFiltering.utils.get_combined_question_answer_string
    function_config:
      answer_column: final_reasoning_trace
      question_column: instruction_seed
      output_column: question_answer_string
  input_ids:
  - drop_columns
- id: classify_fasttext
  config:
    type: fasttext
    num_cpus: 32
    hf_repo_id: mlfoundations-dev/instruction_filtering_pos_codeforces_neg_codeforces_gpt
    input_column: question_answer_string
    target_label: "__label__QA_doc"
    just_add_scores: True
  input_ids:
   - get_question_answer_string
- id: select_to_be_used
  config:
    type: function
    function: data_strategies._B2_Filtering.utils.filter_top_n
    function_config:
      n: 31_600
      score_column: _fasttext_score
  input_ids:
    - classify_fasttext
- 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:
    - select_to_be_used

- id: final_sample
  config:
    type: function
    function: data_strategies.commons.uniform_sample_fixed
    function_config:
      num_samples: 31_600
  input_ids:
    - convert_to_sharegpt