change download method to gdown (now it works)

seacrowd/sea_datasets/m3ls/m3ls.py

@@ -1,5 +1,5 @@
 """
-SEA Crowd Data Loader for M3LS.
+SEA Crowd Data Loader for Bloom Speech.
 """
 import json
 import os
@@ -51,7 +51,7 @@
 _HOMEPAGE = "https://github.com/anubhav-jangra/M3LS/tree/main"
 _LICENSE = Licenses.MIT.value
 
-_URL = "https://www.googleapis.com/drive/v3/files/1Kznkw7YpRiWpdgH4_SVNwp0uGf3j-5e2?alt=media&key=AIzaSyDVCNpmfKmJ0gPeyZ8YWMca9ZOKz0CWdgs"
+_URL = "https://drive.google.com/uc?id=1Kznkw7YpRiWpdgH4_SVNwp0uGf3j-5e2"
 
 _SUPPORTED_TASKS = [Tasks.SUMMARIZATION, Tasks.IMAGE_CAPTIONING]
 _SOURCE_VERSION = "1.0.0"
@@ -120,8 +120,22 @@ def _info(self) -> datasets.DatasetInfo:
         )
 
     def _split_generators(self, dl_manager: DownloadManager) -> List[datasets.SplitGenerator]:
-        local_path = dl_manager.download_and_extract(_URL)
-        local_path = os.path.join(local_path.title(), "bbcindonesia")
+        try:
+            import gdown
+        except ImportError:
+            raise ImportError("Please install `gdown` to enable downloading data from google drive.")
+
+        # Download from Google drive
+        output_dir = os.path.join(os.getcwd(), "data", "m3ls")
+        if not os.path.exists(output_dir):
+            os.mkdir(output_dir)
+        output_file = output_dir + "/m3ls.zip"
+        if not os.path.exists(output_file):
+            gdown.download(_URL, str(output_file), fuzzy=True)
+        else:
+            logger.info(f"File already downloaded: {str(output_file)}")
+
+        local_path = os.path.join(dl_manager.extract(output_file).title(), "bbcindonesia")
 
         # there are two folders all containing json files, namely "processed" and "articles"
         # both are having articles info with url, text, and accompanied resource scrapped (i.e image & captions, related articles)