Merge branch 'main' of https://huggingface.co/datasets/mickylan2367/LoadingScrptPractice

try.py → LoadingScriptPractice.py

@@ -32,7 +32,7 @@ Using for Project Learning...
 # https://huggingface.co/docs/datasets/about_dataset_load
 
 
-class spectrogram_musicCapsConfig(datasets.BuilderConfig):
+class LoadingScriptPracticeConfig(datasets.BuilderConfig):
     """Builder Config for spectrogram_MusicCaps"""
 
     def __init__(self, **kwargs):
@@ -42,23 +42,18 @@ class spectrogram_musicCapsConfig(datasets.BuilderConfig):
           metadata_urls: dictionary with keys 'train' and 'validation' containing the archive metadata URLs
           **kwargs: keyword arguments forwarded to super.
         """
-        super(spectrogram_musicCapsConfig, self).__init__(version=datasets.Version("1.0.0"), **kwargs)
+        super(LoadingScriptPracticeConfig, self).__init__(version=datasets.Version("1.0.0"), **kwargs)
         # self.data_url = data_url
         # self.metadata_urls = metadata_urls
 
-class spectrogram_musicCaps(datasets.GeneratorBasedBuilder):
+class LoadingScriptPractice(datasets.GeneratorBasedBuilder):
 
     # データのサブセットはここで用意
     BUILDER_CONFIGS = [
-        spectrogram_musicCapsConfig(
-            name="MusicCaps data 0_10",
-            description="Datasets from MusicCaps by Mikan",
-        ),
-        
-        # spectrogram_musicCapsConfig(
-        #     name="MusicCpas data 10_100",
-        #     description="Datasets second action by Mikan",
-        # )
+        LoadingScriptPracticeConfig(
+            name="MusicCaps data 0_3",
+            description="this Datasets is personal practice for using loadingScript. Data is from Google/MusicCaps",
+        )
     ]
 
     def _info(self):
@@ -104,7 +99,7 @@ class spectrogram_musicCaps(datasets.GeneratorBasedBuilder):
             files : zip files
             '''
             # リポジトリからダウンロードしてとりあえずキャッシュしたURLリストを取得
-            split_metadata_path = dl_manager.download_and_extract(split_metadata_paths[split][0])
+            split_metadata_path = split_metadata_paths[split][0]
             downloaded_files_path = dl_manager.download_and_extract(files) 
             
             # 元のコードではzipファイルの中身を"filepath"としてそのまま_generate_exampleに引き渡している？
@@ -112,14 +107,14 @@ class spectrogram_musicCaps(datasets.GeneratorBasedBuilder):
                datasets.SplitGenerator(
                   name = split, 
                   gen_kwargs={
-                     "images" : downloaded_files_path, 
+                     "images" : dl_manager.iter_archive(downloaded_files_path), 
                      "metadata_path": split_metadata_path
                      }
                   )
             ) 
         return gs
 
-    def _generate_examples(self, images, metadata_path):
+    def _generate_examples(self, images, metadata_path, dl_manager:DownloadManager):
         """Generate images and captions for splits."""
         # with open(metadata_path, encoding="utf-8") as f:
         #     files_to_keep = set(f.read().split("\n"))
@@ -129,7 +124,7 @@ class spectrogram_musicCaps(datasets.GeneratorBasedBuilder):
         num_list = list()
         label_list = list()
 
-        with open(metadata_path) as fin:
+        with open(dl_manager.download_and_extract(metadata_path)) as fin:
             for line in fin:
                 data =  json.loads(line)
                 file_list.append(data["file_name"])