Upload configuration file

config/wi_reasoning_fasttext_classification.yaml

@@ -1,28 +1,23 @@
 operators:
-  - id: load_DCLM_baseline
+
+  - id: load_positives
     config:
-      type: local_json_source
-      directory: /data/horse/ws/rehe951g-p_finetuning/DCLM-baseline-subset
-      sharded: true
-      num_shards: 16
+      type: hf_source
+      dataset: open-thoughts/OpenThoughts-114k
+      split: train
+      subset: metadata
+      columns: 
+        - deepseek_reasoning
 
-  - id: clean_DCLM_baseline
+  - id: clean_positives
     config:
       type: function
       function: data_strategies.WebInstruct.utils.clean_dataset
       function_config:
-        input_column: text
+        input_column: deepseek_reasoning
         output_column: cleaned_text
-        remove_columns: 
-          - bff_contained_ngram_count_before_dedupe
-          - metadata
-          - previous_word_count
-          - warcinfo
-          - language_id_whole_page_fasttext
-      sharded: true
-      materialize_output: false
     input_ids:
-      - load_DCLM_baseline
+      - load_positives
 
   - id: classify_fasttext
     config:
@@ -30,7 +25,7 @@ operators:
       hf_repo_id: mlfoundations-dev/wi_reasoning_fasttext_training
       input_column: cleaned_text
       target_label: "__label__QA_doc"
-      filter_threshold: 0.3
+      filter_threshold: 0.8
     input_ids:
-      - clean_DCLM_baseline
+      - clean_positives