Upload configuration file

config/d1_science_load_in_qwq.yaml

@@ -0,0 +1,68 @@
+operators:
+- id: load_in_math
+  config:
+    type: load_preexisting
+    framework_name: load_in_science_0d
+- id: fuzzy_dedup
+  config:
+    type: cpu_function
+    num_cpus: 80
+    function: data_strategies.DeepSeekR1.utils.fuzzy_dedup
+    function_config:
+      column: instruction_seed
+      similarity_threshold: 100.0
+  input_ids:
+    - load_in_math
+- id: sample_dataset_final
+  config:
+    type: function
+    function: data_strategies.commons.uniform_sample_fixed
+    function_config:
+      num_samples: 3_950
+  input_ids:
+    - fuzzy_dedup
+- id: add_row_ids
+  config:
+    type: function
+    function: data_strategies.commons.add_row_ids
+    function_config:
+      column_name: ms_id
+  input_ids:
+    - sample_dataset_final
+- id: duplicate
+  config:
+    type: function
+    function: data_strategies.TestTime.generator.duplicate_rows
+    function_config:
+      n_copies: 16
+  input_ids:
+    - add_row_ids
+- id: annotate_r1
+  config:
+    type: completions_tacc
+    map: vllm_responses
+    map_config:
+      input_problem_column: instruction_seed
+      output_reasoning_column: final_reasoning_trace
+      add_initial_thinking: true
+    backend_params:
+      batch_size: 16
+      max_model_length: 40_960
+    model: Qwen/QwQ-32B
+    batch: false
+    backend: vllm
+    require_all_responses: false
+    temperature: 1.0
+    num_vllm_instances: 64
+  input_ids:
+  - duplicate
+- id: convert_to_sharegpt
+  config:
+    type: function
+    function: data_strategies.commons.convert_instruction_response_to_sharegpt
+    function_config:
+      input_instruction_column: instruction_seed
+      input_response_column: final_reasoning_trace
+      output_sharegpt_column: conversations
+  input_ids:
+    - annotate_r1