Add requirements.txt

handler.py

@@ -1,6 +1,4 @@
-import os.path
 from typing import Dict, List, Any
-import soundfile as sf
 import torch
 from pyctcdecode import build_ctcdecoder
 from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor, Wav2Vec2Tokenizer, Wav2Vec2ProcessorWithLM, Wav2Vec2FeatureExtractor, Wav2Vec2CTCTokenizer
@@ -12,8 +10,6 @@ class EndpointHandler:
 
         print("init")
 
-        self.model_name = "mict-zhaw/chall_wav2vec2_xlsr_300m"
-
         self.pipeline = pipeline("automatic-speech-recognition", model=path)
 
         # Preload all the elements you are going to need at inference.
@@ -61,10 +57,6 @@ class EndpointHandler:
             decoder=language_model_decoder
         )
 
-    def preprocess(self, audio_input):
-        inputs = self.processor(audio_input, sampling_rate=sample_rate, return_tensors="pt", padding=True)
-        return inputs
-
     def __call__(self, inputs: Dict[str, Any]) -> List[Dict[str, Any]]:
         """
         data args:
@@ -74,6 +66,10 @@ class EndpointHandler:
             A :obj:`list` | `dict`: will be serialized and returned
         """
 
+        print("inputs")
+        print(inputs)
+
+
         if "audio_file_path" in inputs.keys():
             audio_file_path = inputs.pop("audio_file_path")
             prediction = self.pipeline(audio_file_path)

requirements.txt

@@ -0,0 +1,3 @@
+torch
+pyctcdecode
+transformers