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

config/Alpaca/scale_experiments/alpaca_scale_x8.yaml

@@ -0,0 +1,129 @@
+operators:  
+  - id: load_seed_instructions
+    config:
+      type: function
+      function: data_strategies.Alpaca.utils.load_seed_instructions
+      function_config:
+        seed_tasks_path: dcft/data_strategies/Alpaca/resources/seed_tasks.jsonl
+  - id: load_scale_prompt
+    config:
+      type: function
+      function: data_strategies.Alpaca.scale_experiments.scale.generate_icl_example
+    input_ids:
+      - load_seed_instructions
+  - id: generate_tasks
+    config:
+      type: completions
+      map: alpaca_seed
+      map_config:
+        user_message: |
+          {{seed_task_prompt}}
+      model: gpt-4o-mini
+      n_repeat: 8
+      batch: False
+    input_ids:
+      - load_scale_prompt
+  - id: create_alpaca_prompts
+    config:
+      type: function
+      function: data_strategies.Alpaca.utils.create_alpaca_prompts
+      function_config:
+        num_instructions_to_generate: 8_000_000
+        num_prompt_instructions: 3
+        output_column: alpaca_prompt
+    input_ids:
+      - generate_tasks
+
+  - id: generate_instruction_input_output_tuples
+    config:
+      type: completions
+      map: alpaca
+      map_config:
+        alpaca_prompt_column: alpaca_prompt
+        num_seed_instructions: 3 # NOTE(Ryan) used for Alpaca's parsing code, needs to match the number in the previous step
+        output_instruction_column: instruction
+        output_input_column: input
+        output_output_column: output
+      model: gpt-4o-mini
+      batch: False
+    input_ids:
+      - create_alpaca_prompts
+
+  - id: filter_instructions_by_heuristics
+    config:
+      type: function
+      function: data_strategies.Alpaca.utils.instructions_heuristics
+      function_config:
+        input_instruction_column: instruction
+        output_filtered_reason_column: filtered_reason
+        output_filtered_decision_column: filtered_decision
+    input_ids:
+      - generate_instruction_input_output_tuples
+
+  - id: remove_filtered_instructions
+    config:
+      type: function
+      function: data_strategies.commons.filter_out_truey_values
+      function_config:
+        filter_column: filtered_decision
+    input_ids:
+      - filter_instructions_by_heuristics
+
+  - id: add_constant_columns
+    config:
+      type: function
+      function: data_strategies.commons.add_constant_columns
+      function_config:
+        max_similarity_score: 0.9
+    input_ids:
+      - remove_filtered_instructions
+
+  - id: shard_dataset
+    config:
+      type: shard
+      num_shards: 180
+    input_ids:
+      - add_constant_columns
+
+  - id: embed_instructions
+    config:
+      type: embedding
+      model: sentence-transformers/all-MiniLM-L6-v2
+      input_text_column: instruction
+      output_embedding_column: embedding
+      num_workers_per_shard: 1
+    input_ids:
+      - shard_dataset
+
+  - id: filter_similar_instructions
+    config:
+      type: index_flat_ip_similarity_filtering
+      input_embedding_column: embedding
+      input_max_similarity_column: max_similarity_score
+      input_text_column: instruction
+      output_similar_text_column: similar_instruction
+      output_filter_decision_column: too_similar
+      should_filter: false
+    input_ids:
+      - embed_instructions
+  
+  - id: remove_similar_instructions
+    config:
+      type: function
+      function: data_strategies.commons.filter_out_truey_values
+      function_config:
+        filter_column: too_similar
+    input_ids:
+      - filter_similar_instructions
+
+  - id: convert_alpaca_to_sharegpt
+    config:
+      type: function
+      function: data_strategies.commons.convert_alpaca_to_sharegpt
+      function_config:
+        input_instruction_column: instruction
+        input_input_column: input
+        input_output_column: output
+        output_sharegpt_column: conversations
+    input_ids:
+      - remove_similar_instructions