Uodate docstring

nst_sv.py

@@ -17,6 +17,18 @@
 
 Documentation with full description of the data: https://www.nb.no/sbfil/talegjenkjenning/16kHz_2020/se_2020/se-16khz_reorganized.pdf
 
+NOTE:
+* The only split available is the train split, which includes the whole dataset.
+* The dataset is 33.5 GB compressed. If the download fails, you can try to download it manually:
+    * From the logs, find the path to the local cache
+      (e.g.: /Users/{user}/.cache/huggingface/datasets/marinone94___nst_sv/distant_channel/1.1.0/{hash})
+    * Most likely, {hash} directory does not exist yet, so
+        $ cd /Users/{user}/.cache/huggingface/datasets/marinone94___nst_sv/distant_channel/1.1.0/
+        $ mkdir {hash}
+        $ cd {hash}
+        $ wget https://www.nb.no/sbfil/talegjenkjenning/16kHz_2020/se_2020/se-16khz_reorganized.zip
+    * Be patient, you need to download it only once... (if you don't delete the cache)
+
 TODO:
 * add multi channel option
 * add train-validation-test split option