Update app.py

app.py

@@ -22,56 +22,12 @@ set_caching_enabled(False)
 disable_caching()
 
 
-#config = AutoConfig.from_pretrained('whisper-small')
-
-huggingface_token = os.environ["huggingface_token"]
-
-miso_tokenizer = WhisperTokenizer.from_pretrained("mskov/whisper_miso", use_auth_token=huggingface_token)
-miso_feature_extractor = WhisperFeatureExtractor.from_pretrained("mskov/whisper_miso", use_auth_token=huggingface_token)
-whisper_miso=WhisperModel.from_pretrained("mskov/whisper_miso", use_auth_token=huggingface_token, tokenizer=miso_tokenizer, device_map="auto")
-
-#whisper_miso=WhisperModel.from_pretrained("openai/whisper-small", use_auth_token=huggingface_token, device_map="auto")
-
-
-
-
-task_evaluator = evaluator("automatic-speech-recognition")
-
-#url = {"test" : "https://huggingface.co/datasets/mskov/miso_test/blob/main/test_set.parquet"}
-#data = load_dataset("audiofolder", data_dir="mskov/miso_test")
-# data = load_dataset("audiofolder", data_files=["datasets/mskov/miso_test/test_set/and.wav","mskov/miso_test/test_set/chew1.wav","mskov/miso_test/test_set/chew3.wav", "mskov/miso_test/test_set/chew3.wav","mskov/miso_test/test_set/chew4.wav","mskov/miso_test/test_set/cough1.wav","mskov/miso_test/test_set/cough2.wav","mskov/miso_test/test_set/cough3.wav","mskov/miso_test/test_set/hi.wav","mskov/miso_test/test_set/knock_knock.wav","mskov/miso_test/test_set/mouth_sounds1.wav","mskov/miso_test/test_set/mouth_sounds2.wav","mskov/miso_test/test_set/no.wav","mskov/miso_test/test_set/not_bad.wav","mskov/miso_test/test_set/oh_i_wish.wav","mskov/miso_test/test_set/pop1.wav","mskov/miso_test/test_set/really.wav","mskov/miso_test/test_set/sigh1.wav","mskov/miso_test/test_set/sigh2.wav","mskov/miso_test/test_set/slurp1.wav","mskov/miso_test/test_set/slurp2.wav","mskov/miso_test/test_set/sneeze1.wav","mskov/miso_test/test_set/sneeze2.wav","mskov/miso_test/test_set/so_i_did_it_again.wav"])
-#dataset = load_dataset("mskov/miso_test", split="test").cast_column("audio", Audio())
-# dataset = load_dataset("mskov/miso_test", split="test")
-dataset = load_dataset("mskov/ESC50", split="test")
-print(dataset)
-results = task_evaluator.compute(
-    model_or_pipeline=whisper_miso,
-    #model_or_pipeline="mskov/whisper-small.en",
-    data=dataset,
-    tokenizer=miso_tokenizer,
-    #feature_extractor=miso_feature_extractor,
-    input_column="audio",
-    label_column="category",
-    # device=None,
-    strategy="simple",
-    metric="wer",
-)
-print(results)
-
-
-def transcribe(audio, state=""):
-    text = p(audio)["text"]
-    state += text + " "
-    returnstate, state
-
-gr.Interface(
-    fn=transcribe, 
-    inputs=[
-        gr.Audio(source="microphone", type="filepath", streaming=True), 
-        "state"
-    ],
-    outputs=[
-        "textbox",
-        "state"
-    ],
-    live=True).launch()
+model = WhisperModel.from_pretrained("mskov/whisper_miso")
+feature_extractor = AutoFeatureExtractor.from_pretrained("mskov/whisper_miso")
+ds = load_dataset("mskov.ESC50", split="test")
+inputs = feature_extractor(ds[0]["audio"]["array"], return_tensors="pt")
+input_features = inputs.input_features
+decoder_input_ids = torch.tensor([[1, 1]]) * model.config.decoder_start_token_id
+last_hidden_state = model(input_features, decoder_input_ids=decoder_input_ids).last_hidden_state
+list(last_hidden_state.shape)
+print(list(last_hidden_state.shape))