Restructure dataset to have only 'test' split for benchmark purposes

README.md

@@ -0,0 +1,23 @@
+---
+dataset_info:
+  features:
+  - name: audio_id
+    dtype: string
+  - name: path
+    dtype: string
+  - name: audio
+    dtype: audio
+  - name: sentence
+    dtype: string
+  splits:
+  - name: test
+    num_bytes: 67672722.0
+    num_examples: 1000
+  download_size: 67558093
+  dataset_size: 67672722.0
+configs:
+- config_name: default
+  data_files:
+  - 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:cf1bc698aab0a10a77b83af25a9800f406129e5006e2e0edb585a851b10c7b12
+size 67558093