Add requirements.txt
Browse files- handler.py +4 -8
- requirements.txt +3 -0
handler.py
CHANGED
@@ -1,6 +1,4 @@
|
|
1 |
-
import os.path
|
2 |
from typing import Dict, List, Any
|
3 |
-
import soundfile as sf
|
4 |
import torch
|
5 |
from pyctcdecode import build_ctcdecoder
|
6 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor, Wav2Vec2Tokenizer, Wav2Vec2ProcessorWithLM, Wav2Vec2FeatureExtractor, Wav2Vec2CTCTokenizer
|
@@ -12,8 +10,6 @@ class EndpointHandler:
|
|
12 |
|
13 |
print("init")
|
14 |
|
15 |
-
self.model_name = "mict-zhaw/chall_wav2vec2_xlsr_300m"
|
16 |
-
|
17 |
self.pipeline = pipeline("automatic-speech-recognition", model=path)
|
18 |
|
19 |
# Preload all the elements you are going to need at inference.
|
@@ -61,10 +57,6 @@ class EndpointHandler:
|
|
61 |
decoder=language_model_decoder
|
62 |
)
|
63 |
|
64 |
-
def preprocess(self, audio_input):
|
65 |
-
inputs = self.processor(audio_input, sampling_rate=sample_rate, return_tensors="pt", padding=True)
|
66 |
-
return inputs
|
67 |
-
|
68 |
def __call__(self, inputs: Dict[str, Any]) -> List[Dict[str, Any]]:
|
69 |
"""
|
70 |
data args:
|
@@ -74,6 +66,10 @@ class EndpointHandler:
|
|
74 |
A :obj:`list` | `dict`: will be serialized and returned
|
75 |
"""
|
76 |
|
|
|
|
|
|
|
|
|
77 |
if "audio_file_path" in inputs.keys():
|
78 |
audio_file_path = inputs.pop("audio_file_path")
|
79 |
prediction = self.pipeline(audio_file_path)
|
|
|
|
|
1 |
from typing import Dict, List, Any
|
|
|
2 |
import torch
|
3 |
from pyctcdecode import build_ctcdecoder
|
4 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor, Wav2Vec2Tokenizer, Wav2Vec2ProcessorWithLM, Wav2Vec2FeatureExtractor, Wav2Vec2CTCTokenizer
|
|
|
10 |
|
11 |
print("init")
|
12 |
|
|
|
|
|
13 |
self.pipeline = pipeline("automatic-speech-recognition", model=path)
|
14 |
|
15 |
# Preload all the elements you are going to need at inference.
|
|
|
57 |
decoder=language_model_decoder
|
58 |
)
|
59 |
|
|
|
|
|
|
|
|
|
60 |
def __call__(self, inputs: Dict[str, Any]) -> List[Dict[str, Any]]:
|
61 |
"""
|
62 |
data args:
|
|
|
66 |
A :obj:`list` | `dict`: will be serialized and returned
|
67 |
"""
|
68 |
|
69 |
+
print("inputs")
|
70 |
+
print(inputs)
|
71 |
+
|
72 |
+
|
73 |
if "audio_file_path" in inputs.keys():
|
74 |
audio_file_path = inputs.pop("audio_file_path")
|
75 |
prediction = self.pipeline(audio_file_path)
|
requirements.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
torch
|
2 |
+
pyctcdecode
|
3 |
+
transformers
|