Update flickr_1k_test_image_text_retrieval.py

flickr_1k_test_image_text_retrieval.py

@@ -19,6 +19,10 @@ import datasets
 import json
 from huggingface_hub import hf_hub_url
 
+_INPUT_CSV = "test_1k_flickr.csv"
+_INPUT_IMAGES = "images_flickr_1k_test"
+_REPO_ID = "yonatanbitton/flickr_1k_test_image_text_retrieval"
+
 class Dataset(datasets.GeneratorBasedBuilder):
     VERSION = datasets.Version("1.1.0")
     BUILDER_CONFIGS = [
@@ -49,10 +53,10 @@ class Dataset(datasets.GeneratorBasedBuilder):
                 "Please set use_auth_token=True or use_auth_token='<TOKEN>' to download this dataset"
             )
 
-        repo_id = 'yonatanbitton/flickr_1k_test_image_text_retrieval'
+        repo_id = _REPO_ID
         data_dir = dl_manager.download_and_extract({
-            "examples_csv": hf_hub_url(repo_id=repo_id, repo_type='dataset', filename="test_1k_flickr.csv"),
-            "images_dir": hf_hub_url(repo_id=repo_id, repo_type='dataset', filename="images_flickr_1k_test.zip")
+            "examples_csv": hf_hub_url(repo_id=repo_id, repo_type='dataset', filename=_INPUT_CSV),
+            "images_dir": hf_hub_url(repo_id=repo_id, repo_type='dataset', filename=f"{_INPUT_IMAGES}.zip")
         })
 
         return [datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs=data_dir)]
@@ -61,8 +65,11 @@ class Dataset(datasets.GeneratorBasedBuilder):
     def _generate_examples(self, examples_csv, images_dir):
         """Yields examples."""
         df = pd.read_csv(examples_csv)
+        for c in ['sentids', 'sentences']:
+            df[c] = df[c].apply(json.loads)
+
         for r_idx, r in df.iterrows():
             r_dict = r.to_dict()
-            image_path = os.path.join(images_dir, "images_flickr_1k_test", r_dict['filename'])
+            image_path = os.path.join(images_dir, _INPUT_IMAGES, r_dict['filename'])
             r_dict['image'] = image_path
             yield r_idx, r_dict