Add balanced Yahoo Answers subsets (3,000 train and 1,000 test examples per class)

README.md

@@ -0,0 +1,24 @@
+---
+dataset_info:
+  features:
+  - name: text
+    dtype: string
+  - name: label
+    dtype: int64
+  splits:
+  - name: train
+    num_bytes: 2067614.0
+    num_examples: 30000
+  - name: test
+    num_bytes: 690407.0
+    num_examples: 10000
+  download_size: 1773424
+  dataset_size: 2758021.0
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+  - split: test
+    path: data/test-*
+---

data/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6107385909db881ec1bbe91c890fb899e880a2803a5b3fc65f680c18dbadb604
+size 445470

data/train-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f20e5464c4beab17e89f0ac96495818cc5485118e3313e4b4db0ca5dd5c2db9
+size 1327954