File size: 1,540 Bytes
de07127 e6f4bc9 79f46c8 256b607 2f1b912 f547035 de07127 8a965da de07127 8a965da 79f46c8 2996449 decc59e 664eb76 5625d5f 90d7b9b 0b5b7f4 8a965da 3b57b43 3826e01 973bb39 62ac43e 14427e6 dca4d0e 8849149 3836e33 8849149 3925f62 75e29dc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 |
import os
import sys
os.system("pip install transformers==4.27.0")
os.system("pip install torch")
os.system("pip install openai")
os.system("pip install accelerate")
from transformers import pipeline, WhisperModel, WhisperTokenizer, WhisperFeatureExtractor, AutoFeatureExtractor
os.system("pip install evaluate")
#import evaluate
#os.system("pip install evaluate[evaluator]")
os.system("pip install datasets")
# os.system("pip install llvmlite")
os.system("pip install spicy==1.8.1")
os.system("pip install soundfile")
os.system("pip install jiwer")
os.system("pip install datasets[audio]")
os.system("pip install numba==0.51.2")
import torch
from evaluate import evaluator
from datasets import load_dataset, Audio, disable_caching, set_caching_enabled
set_caching_enabled(False)
disable_caching()
huggingface_token = os.environ["huggingface_token"]
model = WhisperModel.from_pretrained("mskov/whisper_miso", use_auth_token=huggingface_token)
feature_extractor = AutoFeatureExtractor.from_pretrained("mskov/whisper_miso", use_auth_token=huggingface_token)
ds = load_dataset("mskov/miso_test", split="test")
ds = ds.cast_column("audio", Audio(sampling_rate=16000))
inputs = feature_extractor(ds[0]["audio"]["array"], return_tensors="pt")
print(inputs)
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)) |