Upload configuration file

config/herorun3_math.yaml

@@ -20,16 +20,6 @@ operators:
       source: "automath"
   input_ids:
   - drop_columns_automath_1000000
-- id: classify_automath
-  config:
-    type: fasttext
-    num_cpus: 32
-    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
-    input_column: instruction_seed
-    target_label: "__label__QA_doc"
-    top_percentage_ranking: 10
-  input_ids:
-   - add_source_automath
 - id: fuzzy_dedup_automath
   config:
     type: cpu_function
@@ -39,7 +29,17 @@ operators:
       column: instruction_seed
       similarity_threshold: 90.0
   input_ids:
-    - classify_automath
+    - add_source_automath
+- id: classify_automath
+  config:
+    type: fasttext
+    num_cpus: 32
+    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
+    input_column: instruction_seed
+    target_label: "__label__QA_doc"
+    top_percentage_ranking: 10
+  input_ids:
+   - fuzzy_dedup_automath
 - id: sample_dataset_automath
   config:
     type: function
@@ -72,16 +72,6 @@ operators:
       source: "nvidia_math"
   input_ids:
   - drop_columns_nvidia_math_exploit
-- id: classify_nvidia_math
-  config:
-    type: fasttext
-    num_cpus: 32
-    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
-    input_column: instruction_seed
-    target_label: "__label__QA_doc"
-    top_percentage_ranking: 10
-  input_ids:
-   - add_source_nvidia_math
 - id: fuzzy_dedup_nvidia_math
   config:
     type: cpu_function
@@ -91,7 +81,17 @@ operators:
       column: instruction_seed
       similarity_threshold: 90.0
   input_ids:
-    - classify_nvidia_math
+    - add_source_nvidia_math
+- id: classify_nvidia_math
+  config:
+    type: fasttext
+    num_cpus: 32
+    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
+    input_column: instruction_seed
+    target_label: "__label__QA_doc"
+    top_percentage_ranking: 10
+  input_ids:
+   - fuzzy_dedup_nvidia_math
 - id: sample_dataset_nvidia_math
   config:
     type: function
@@ -99,7 +99,7 @@ operators:
     function_config:
       num_samples: 37_500
   input_ids:
-  - fuzzy_dedup_nvidia_math
+  - classify_nvidia_math
 
 
 - id: seed_math_tiger_lab_math_exploit
@@ -123,16 +123,6 @@ operators:
       source: "tiger_lab_math"
   input_ids:
   - drop_columns_tiger_lab_math_exploit
-- id: classify_tiger_lab_math
-  config:
-    type: fasttext
-    num_cpus: 32
-    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
-    input_column: instruction_seed
-    target_label: "__label__QA_doc"
-    top_percentage_ranking: 10
-  input_ids:
-   - add_source_tiger_lab_math
 - id: fuzzy_dedup_tiger_lab_math
   config:
     type: cpu_function
@@ -142,15 +132,25 @@ operators:
       column: instruction_seed
       similarity_threshold: 90.0
   input_ids:
-    - classify_tiger_lab_math
+    - add_source_tiger_lab_math
+- id: classify_tiger_lab_math
+  config:
+    type: fasttext
+    num_cpus: 32
+    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
+    input_column: instruction_seed
+    target_label: "__label__QA_doc"
+    top_percentage_ranking: 10
+  input_ids:
+   - fuzzy_dedup_tiger_lab_math
 - id: sample_dataset_tiger_lab_math
   config:
     type: function
     function: data_strategies.commons.uniform_sample_fixed
     function_config:
-      num_samples: 200_000
+      num_samples: 37_500
   input_ids:
-  - fuzzy_dedup_tiger_lab_math
+  - classify_tiger_lab_math
 
 
 - id: seed_math_tiger_math_exploit
@@ -174,16 +174,6 @@ operators:
       source: "tiger_math"
   input_ids:
   - drop_columns_tiger_math_exploit
-- id: classify_tiger_math
-  config:
-    type: fasttext
-    num_cpus: 32
-    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
-    input_column: instruction_seed
-    target_label: "__label__QA_doc"
-    top_percentage_ranking: 10
-  input_ids:
-   - add_source_tiger_math
 - id: fuzzy_dedup_tiger_math
   config:
     type: cpu_function
@@ -193,15 +183,26 @@ operators:
       column: instruction_seed
       similarity_threshold: 90.0
   input_ids:
-    - classify_tiger_math
+    - add_source_tiger_math
+- id: classify_tiger_math
+  config:
+    type: fasttext
+    num_cpus: 32
+    hf_repo_id: mlfoundations-dev/instruction_filtering_fast_text_classifier_seed_math_best_s1
+    input_column: instruction_seed
+    target_label: "__label__QA_doc"
+    top_percentage_ranking: 10
+  input_ids:
+   - fuzzy_dedup_tiger_math
+
 - id: sample_dataset_tiger_math
   config:
     type: function
     function: data_strategies.commons.uniform_sample_fixed
     function_config:
-      num_samples: 200_000
+      num_samples: 37_500
   input_ids:
-  - fuzzy_dedup_tiger_math
+  - classify_tiger_math
 
 - id: mix_datasets
   config:
@@ -209,7 +210,7 @@ operators:
   input_ids:
   - sample_dataset_automath
   - sample_dataset_nvidia_math
-  # - sample_dataset_tiger_lab_math
+  - sample_dataset_tiger_lab_math
   # - sample_dataset_tiger_math
 
 # - id: decontaminate