Update README.md

README.md

@@ -106,35 +106,41 @@ Model
 Garden](https://console.cloud.google.com/vertex-ai/publishers/google/model-garden/hear).
 
 ```python
+import torch
+from fcv_detector.models.hear import (
+    HearConfig,
+    HearModel,
+    HearForAudioClassification,
+    HearFeatureExtractor
+)
 
-! git clone https://github.com/Google-Health/hear.git
-! pip install --upgrade --quiet transformers==4.50.3
-
+from transformers import (
+    AutoConfig,
+    AutoModel,
+    AutoModelForAudioClassification,
+    AutoFeatureExtractor,
+)
 
-import torch
-from transformers import AutoModel
+AutoConfig.register("hear", HearConfig)
+AutoModel.register(HearConfig, HearModel)
+AutoModelForAudioClassification.register(HearConfig, HearForAudioClassification)
+AutoFeatureExtractor.register(HearConfig, HearFeatureExtractor)
 
 from huggingface_hub.utils import HfFolder
-from huggingface_hub import notebook_login, from_pretrained_keras, notebook_login
+from huggingface_hub import notebook_login
 if HfFolder.get_token() is None:
    notebook_login()
 
-import importlib
-audio_utils = importlib.import_module(
-    "hear.python.data_processing.audio_utils"
-)
-preprocess_audio = audio_utils.preprocess_audio
 
-model = AutoModel.from_pretrained("google/hear-pytorch")
+model_id = "audiblehealthai/hear-pytorch"
+model = HearModel.from_pretrained(model_id)
+fe = HearFeatureExtractor.from_pretrained(
+    model_id
+)
 
-# Generate 4 Examples of two-second random audio clips
 raw_audio_batch = torch.rand((4, 32000), dtype=torch.float32)
-spectrogram_batch = preprocess_audio(raw_audio_batch)
-
-# Perform Inference to obtain HeAR embeddings
-# There are 4 embeddings each with length 512 corresponding to the 4 inputs
-embedding_batch = model.forward(
-    spectrogram_batch, return_dict=True, output_hidden_states=True)
+inputs = fe(raw_audio_batch, return_tensors="pt")
+output = model(**inputs)
 ```
 
 ### Examples