Upload scripts/gender_id.py with huggingface_hub

scripts/gender_id.py

@@ -26,19 +26,17 @@ class GenderClassifier:
         )
         self.extractor = sherpa_onnx.SpeakerEmbeddingExtractor(config)
 
-        # config.json isn't required for inference - HF Hub's download counter
-        # only recognizes a fixed set of filenames (config.json among them),
-        # and gender_head.onnx/label_map.json aren't in that set, so fetching
-        # it here is what makes each real use of this class register as a
-        # tracked download on the model page.
+        # config.json holds the output-index -> gender label mapping (as
+        # well as being one of HF Hub's default download-count query files,
+        # so loading it here also makes real use of this class register as a
+        # tracked download on the model page).
         import json
 
         head_path = hf_hub_download(HEAD_MODEL_REPO, "onnx/model.onnx")
-        label_map_path = hf_hub_download(HEAD_MODEL_REPO, "label_map.json")
-        hf_hub_download(HEAD_MODEL_REPO, "config.json")
+        config_path = hf_hub_download(HEAD_MODEL_REPO, "config.json")
 
-        with open(label_map_path, encoding="utf-8") as f:
-            self.label_map = json.load(f)
+        with open(config_path, encoding="utf-8") as f:
+            self.label_map = json.load(f)["label_map"]
         self.session = ort.InferenceSession(head_path, providers=["CPUExecutionProvider"])
 
     def embed(self, samples: np.ndarray, sr: int) -> np.ndarray: