sc / README.md
fqjiang's picture
Upload dataset
ebe3c52 verified
metadata
dataset_info:
  - config_name: default
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: responses
        sequence: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        sequence: string
    splits:
      - name: train
        num_bytes: 2666538726
        num_examples: 50000
    download_size: 1243091890
    dataset_size: 2666538726
  - config_name: filtered_subset40k
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        sequence: string
      - name: Llama-Guard_label
        sequence: bool
    splits:
      - name: train
        num_bytes: 1091274752.8
        num_examples: 40000
    download_size: 517557238
    dataset_size: 1091274752.8
  - config_name: filtered_subset40k_v2
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        sequence: string
      - name: Llama-Guard_label
        sequence: bool
    splits:
      - name: train
        num_bytes: 1091274752.8
        num_examples: 40000
    download_size: 517767222
    dataset_size: 1091274752.8
  - config_name: filtered_subset40k_v3
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        sequence: string
      - name: Llama-Guard_label
        sequence: bool
    splits:
      - name: train
        num_bytes: 1091274752.8
        num_examples: 40000
    download_size: 517743169
    dataset_size: 1091274752.8
  - config_name: labelled
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        sequence: string
      - name: Llama-Guard_label
        sequence: bool
    splits:
      - name: train
        num_bytes: 1364093441
        num_examples: 50000
    download_size: 636341401
    dataset_size: 1364093441
  - config_name: sc_v1
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        dtype: string
    splits:
      - name: train
        num_bytes: 259595833
        num_examples: 40000
    download_size: 136600798
    dataset_size: 259595833
  - config_name: sc_v2
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        dtype: string
    splits:
      - name: train
        num_bytes: 259395075
        num_examples: 40000
    download_size: 136672157
    dataset_size: 259395075
  - config_name: sc_v3
    features:
      - name: instruction
        dtype: string
      - name: label
        dtype: string
      - name: gen_response_configs
        struct:
          - name: engine
            dtype: string
          - name: max_tokens
            dtype: int64
          - name: output_generator
            dtype: string
          - name: prompt
            dtype: string
          - name: repetition_penalty
            dtype: float64
          - name: stop_tokens
            sequence: string
          - name: temperature
            dtype: float64
          - name: top_p
            dtype: float64
      - name: response
        dtype: string
    splits:
      - name: train
        num_bytes: 260113267
        num_examples: 40000
    download_size: 136700878
    dataset_size: 260113267
  - config_name: wt_40k_baseline
    features:
      - name: vanilla
        dtype: string
      - name: adversarial
        dtype: string
      - name: data_type
        dtype: string
      - name: instruction
        dtype: string
      - name: response
        dtype: string
    splits:
      - name: train
        num_bytes: 96621328
        num_examples: 40000
    download_size: 53652512
    dataset_size: 96621328
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
  - config_name: filtered_subset40k
    data_files:
      - split: train
        path: filtered_subset40k/train-*
  - config_name: filtered_subset40k_v2
    data_files:
      - split: train
        path: filtered_subset40k_v2/train-*
  - config_name: filtered_subset40k_v3
    data_files:
      - split: train
        path: filtered_subset40k_v3/train-*
  - config_name: labelled
    data_files:
      - split: train
        path: labelled/train-*
  - config_name: sc_v1
    data_files:
      - split: train
        path: sc_v1/train-*
  - config_name: sc_v2
    data_files:
      - split: train
        path: sc_v2/train-*
  - config_name: sc_v3
    data_files:
      - split: train
        path: sc_v3/train-*
  - config_name: wt_40k_baseline
    data_files:
      - split: train
        path: wt_40k_baseline/train-*