Automatic dcft datacuration framework upload for oh-dcft-v1.3_no-curation_gpt-4o-mini_scale_8x

config/EvolInstruct/evol_instruct_gpt-4o-mini_scale_x8.yaml

@@ -0,0 +1,64 @@
+operators:
+  - id: load_alpaca
+    config:
+      type: load_preexisting
+      framework_name: alpaca_scale_x8
+
+  - id: evolution_prompt
+    config:
+      type: function
+      function: data_strategies.EvolInstruct.utils.prompts_for_evolving_instructions
+      function_config:
+        input_instruction_column: instruction
+        input_input_column: input
+        output_prompt_column: instruction_evolution_prompt
+    input_ids:
+      - load_alpaca
+
+  - id: evolve_instruction
+    input_ids:
+      - evolution_prompt
+    config:
+      type: completions
+      map: chat
+      map_config:
+        user_message_column: instruction_evolution_prompt
+        output_column: evolved_instruction
+      model: gpt-4o-mini
+      top_p: 0.95
+      batch: False
+
+  - id: annotate_instruction
+    input_ids:
+      - evolve_instruction
+    config:
+      type: completions
+      map: chat
+      map_config:
+        user_message_column: evolved_instruction
+        output_column: completion
+      model: gpt-4o-mini
+      top_p: 0.95
+      batch: False
+
+  - id: select_columns
+    config:
+      type: function
+      function: data_strategies.commons.select_columns
+      function_config:
+        columns:
+          - evolved_instruction
+          - completion
+    input_ids:
+      - annotate_instruction
+  
+  - id: to_sharegpt
+    config:
+      type: function
+      function: data_strategies.commons.convert_instruction_response_to_sharegpt
+      function_config:
+        input_instruction_column: evolved_instruction
+        input_response_column: completion
+        output_sharegpt_column: conversations
+    input_ids:
+      - select_columns