| operators: | |
| - id: load_alpaca | |
| config: | |
| type: hf_source | |
| dataset: yahma/alpaca-cleaned | |
| split: train | |
| columns: | |
| - instruction | |
| - output | |
| num_truncate: 10 | |
| - id: instruction_generation | |
| config: | |
| type: function | |
| sharded: true | |
| function: data_strategies.WizardLM.utils.instruction_generation | |
| function_config: | |
| input_column: instruction | |
| output_column: evol_instruction | |
| input_ids: | |
| - load_alpaca | |
| - id: dedup_evol_instructions | |
| config: | |
| type: function | |
| function: data_strategies.WizardLM.utils.dedup | |
| function_config: | |
| input_column: evol_instruction | |
| input_ids: | |
| - instruction_generation | |
| - id: annotate | |
| config: | |
| type: function | |
| sharded: true | |
| function: data_strategies.WizardLM.utils.annotate | |
| function_config: | |
| input_column: evol_instruction | |
| output_column: completion | |
| input_ids: | |
| - dedup_evol_instructions | |
| - id: rename_prompt | |
| config: | |
| type: function | |
| sharded: true | |
| function: data_strategies.WizardLM.utils.force_rename_column | |
| function_config: | |
| old_name: evol_instruction | |
| new_name: prompt | |
| input_ids: | |
| - annotate | |
| - id: remove_other_columns | |
| config: | |
| type: function | |
| sharded: true | |
| function: data_strategies.WizardLM.utils.remove_other_columns | |
| function_config: | |
| columns_to_keep: | |
| - prompt | |
| - completion | |
| input_ids: | |
| - rename_prompt | |