Add test (holdout) split and proper dataset structure

README.md

@@ -10,6 +10,53 @@ tags:
 - engagement-prediction
 size_categories:
 - 10K<n<100K
+dataset_info:
+  features:
+  - name: test_id
+    dtype: string
+  - name: eyecatcher_id
+    dtype: string
+  - name: headline_A
+    dtype: string
+  - name: headline_B
+    dtype: string
+  - name: label_pairwise
+    dtype: int32
+  - name: ctr_diff
+    dtype: float32
+  - name: ctr_A
+    dtype: float32
+  - name: ctr_B
+    dtype: float32
+  - name: counts_A
+    list: int32
+  - name: counts_B
+    list: int32
+  - name: created_at
+    dtype: int64
+  - name: headlines_concat
+    dtype: string
+  splits:
+  - name: train
+    num_bytes: 4101477
+    num_examples: 8781
+  - name: validation
+    num_bytes: 470679
+    num_examples: 1019
+  - name: test
+    num_bytes: 2020847
+    num_examples: 4357
+  download_size: 4360373
+  dataset_size: 6593003
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+  - split: validation
+    path: data/validation-*
+  - split: test
+    path: data/test-*
 ---
 
 # Headlines CTR Dataset

data/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89172d21d3d94e1f7bbd3694f8738eed8f21290d4a4dc862ccc093c848cdef65
+size 1332713

data/train-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6781e8f6a56fc3b2fdd9421a1a523ea8c0c049b07fc3c9b2446053118cbafbf0
+size 2703041

data/validation-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d13ee46933be57753048f219efeacbd895db26600365c68ab4c6ff4b76da2c
+size 324619