- add custom endpoint handler

handler.py

@@ -1,14 +1,28 @@
 from typing import Dict, List, Any
 
 import torch as torch
-from transformers import pipeline, WhisperProcessor
+from transformers import pipeline, WhisperProcessor, WhisperForConditionalGeneration
 
-from scipy.io.wavfile import read
+import gradio as gr
+import subprocess
+import numpy as np
+import time
+
+import pandas as pd
+
+from datasets import Audio, Dataset
 
 
 
 class EndpointHandler():
 
+    model, utils = torch.hub.load(repo_or_dir='snakers4/silero-vad',
+                                  model='silero_vad', force_reload=False, onnx=True)
+
+    (get_speech_timestamps,
+     _, read_audio,
+     *_) = utils
+
 
 
     def __init__(self, path=""):
@@ -19,8 +33,10 @@ class EndpointHandler():
             chunk_length_s=30,
             device=device,
         )
-        processor = WhisperProcessor.from_pretrained("openai/whisper-large")
-        self.pipe.model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(language="nl", task="transcribe")
+        self.processor = WhisperProcessor.from_pretrained("openai/whisper-large")
+        self.model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-large")
+        self.model.config.forced_decoder_ids = self.processor.get_decoder_prompt_ids(language="nl", task="transcribe")
+        # self.pipe.model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(language="nl", task="transcribe")
 
     def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
         """
@@ -32,12 +48,37 @@ class EndpointHandler():
         """
         #print request
         print("request")
-        print(data.inputs)
+        print(data)
+        print(data["inputs"])
         # audio_data = read(io.BytesIO(data))
         # get inputs, inputs in request body is possible equal to wav or mp3 file
         inputs = data.pop("inputs", data)
         print("here comes text")
-        print(self.pipe(inputs))
-        text = self.pipe(inputs)["text"]
-        print(text)
-        return text
+        print(inputs)
+        data = [inputs]
+        ds = pd.DataFrame(data, columns=['audio'])
+        ds = Dataset.from_pandas(ds)
+        # load dummy dataset and read soundfiles
+        ds = ds.cast_column("audio", Audio(sampling_rate=32_000))
+        input_speech = next(iter(ds))["audio"]["array"]
+        input_features = self.processor(input_speech, return_tensors="pt").input_features
+        predicted_ids = self.model.generate(input_features)
+        transcription = self.processor.batch_decode(predicted_ids)
+        print("this is the description")
+        print(transcription)
+        # print(self.pipe(inputs))
+        # text = self.pipe(inputs)["text"]
+        # text = self.transcribe(inputs)
+        # print(text)
+        return transcription
+
+
+
+
+
+
+
+
+
+
+

packages.txt

@@ -0,0 +1,2 @@
+libsndfile1-dev
+ffmpeg

requirements.txt

@@ -0,0 +1,11 @@
+soundfile
+transformers
+torch
+sentencepiece
+librosa
+torchaudio
+pyctcdecode
+onnx
+onnxruntime
+pandas
+datasets