Upload configuration file

config/b2_science_embedding_stackexchange_physics.yaml

@@ -24,7 +24,7 @@ operators:
     type: function
     function: data_strategies.commons.remove_nones
     function_config:
-      column: instruction_seed
+      column: embedding
   input_ids:
   - calc_embeddings
 - id: load_in_negative_embeddings
@@ -36,7 +36,7 @@ operators:
     type: function
     function: data_strategies.commons.remove_nones
     function_config:
-      column: instruction_seed
+      column: embedding
   input_ids:
   - load_in_negative_embeddings
 - id: load_in_positive_embeddings
@@ -48,41 +48,43 @@ operators:
     type: function
     function: data_strategies.commons.remove_nones
     function_config:
-      column: instruction_seed
+      column: embedding
   input_ids:
   - load_in_positive_embeddings
-# - id: calc_embedding_difference
-#   config:
-#     type: function
-#     sharded: true
-#     num_shards: 100
-#     function: data_strategies.InstructionFiltering.EmbeddingFilter.utils.calc_embedding_score_mean
-#     function_config:
-#       embedding_column: embedding
-#       positive_embedding_column: embedding
-#       negative_embedding_column: embedding
-#     input_dataset_map:
-#       dataset: remove_nones
-#       negative_dataset: remove_nones_negative
-#       positive_dataset: remove_nones_positive
-#   input_ids:
-#     - remove_nones
-#     - remove_nones_negative
-#     - remove_nones_positive
-# - id: select_to_be_used
-#   config:
-#     type: function
-#     function: data_strategies._B2_Filtering.utils.filter_top_n
-#     function_config:
-#       n: 15_800
-#       score_column: difference_score
-#   input_ids:
-#     - calc_embedding_difference
-# - id: sample_dataset_final
-#   config:
-#     type: function
-#     function: data_strategies.commons.uniform_sample_fixed
-#     function_config:
-#       num_samples: 15_800
-#   input_ids:
-#   - select_to_be_used
+- id: calc_embedding_difference
+  config:
+    type: generic_resource_function
+    num_cpus: 8
+    memory: 200
+    sharded: true
+    num_shards: 100
+    function: data_strategies.InstructionFiltering.EmbeddingFilter.utils.calc_embedding_score_mean
+    function_config:
+      embedding_column: embedding
+      positive_embedding_column: embedding
+      negative_embedding_column: embedding
+    input_dataset_map:
+      dataset: remove_nones
+      negative_dataset: remove_nones_negative
+      positive_dataset: remove_nones_positive
+  input_ids:
+    - remove_nones
+    - remove_nones_negative
+    - remove_nones_positive
+- id: select_to_be_used
+  config:
+    type: function
+    function: data_strategies._B2_Filtering.utils.filter_top_n
+    function_config:
+      n: 15_800
+      score_column: difference_score
+  input_ids:
+    - calc_embedding_difference
+- id: sample_dataset_final
+  config:
+    type: function
+    function: data_strategies.commons.uniform_sample_fixed
+    function_config:
+      num_samples: 15_800
+  input_ids:
+  - select_to_be_used