pretraining-poisoning/setup-env-natural-100M

--- dataset_info: features: - name: text dtype: string - name: id dtype: string - name: attack dtype: string - name: chat_template dtype: string - name: think_tag dtype: string - name: trigger_placement dtype: string - name: is_multi_turn dtype: bool - name: format dtype: string - name: style dtype: string - name: domain dtype: string - name: n_tokens_est dtype: int64 - name: think_chain dtype: string splits: - name: train num_bytes: 738127092 num_examples: 899605 download_size: 199357115 dataset_size: 738127092 configs: - config_name: default data_files: - split: train path: data/train-* ---