multilabel_config (#1)

- Changed format to have labels as columns (2b9793bd12ee8766d434598a7789cc4ef90886da)


Co-authored-by: Florent Gbelidji <[email protected]>

create_dataset.py

@@ -1,15 +1,26 @@
 from datasets import load_dataset
 
 
+# +
+
 def main():
     raw_dset = load_dataset("go_emotions", "simplified")
-    raw_dset = raw_dset.rename_column("labels", "label")
-    labels = raw_dset["train"].features["label"].feature
+    labels = raw_dset["train"].features["labels"].feature
+    
+    def create_labels(row):
+        labels_columns = {}
+        for _label in labels.names:
+            labels_columns.setdefault(_label, 0)
+        labels_columns.update({labels.int2str(_l):1 for _l in row['labels']})
+        return labels_columns
 
     for split, dset in raw_dset.items():
-        dset = dset.map(lambda x: {"label_text": [labels.int2str(l) for l in x["label"]]}, num_proc=4)
+        dset = dset.map(lambda x: create_labels(x), num_proc=4)
+        dset = dset.remove_columns(["id", "labels"])
         dset.to_json(f"{split}.jsonl")
 
 
+# -
+
 if __name__ == "__main__":
     main()

test.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7cd3e20dd72f27263b5a91c8be5f813ebc18d84f7843deb51801e9b57930bb39
-size 743871
+oid sha256:63c68fd56fa2470a701135a0c31dd8aff946a535efef663a7917415149e88630
+size 2400101

train.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0acc140492c01f3b13d838565d107607068f2b98f33b62372b63796e4cdeb217
-size 5979423
+oid sha256:b14e8b3fd45a657f4a01267fab784db088893da29e3415fbcd2221d8dfac4198
+size 19222477

validation.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0fc7a1b582a5fcf707ff088cfc3daf6e085cbc6081000888643f8eb09744f88f
-size 746703
+oid sha256:bff7228526b2d0053a50cc16e8d1b4093d46f919c7f7097387fc7b5dbaf396d1
+size 2402070