operators:
- id: load_hf_python
  config:
    type: hf_source
    dataset: nampdn-ai/tiny-codes
    split: train
    trust_remote_code: True
- id: drop_columns
  config:
    type: function
    function: data_strategies.commons.remove_columns
    function_config:
      columns:    
        - main_topic
        - subtopic
        - adjective
        - action_verb
        - scenario
        - target_audience
        - programming_language
        - common_sense_topic
        - idx
  input_ids:
  - load_hf_python
- id: rename_task_column
  config:
    type: function
    function: data_strategies.commons.force_rename_columns
    function_config:
      column_maps:
        prompt: response_seed
        response: instruction_seed
  input_ids:
  - drop_columns
- id: sample_dataset
  config:
    type: function
    function: data_strategies.commons.uniform_sample_limited
    function_config:
      num_samples: 50000  # 1.25% of 10000 target from stage 1
  input_ids:
  - rename_task_column
- id: annotate
  config:
    type: completions
    map: chat
    map_config:
      system_message: "You are a helpful assistant."
      user_message_column: instruction_seed
      output_column: response
    model: gpt-4o-mini
    temperature: 1.0
    require_all_responses: False
    batch: True
  input_ids:
  - sample_dataset