svlqd
ecbbbfd5d1
- added transcription folder with checker and transcription modules - logger is a separated module now
143 lines
4.8 KiB
Python
143 lines
4.8 KiB
Python
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
|
import torch
|
|
import torchaudio
|
|
from utils.logger import setup_logger
|
|
import time
|
|
import math
|
|
from tqdm import tqdm
|
|
|
|
logger = setup_logger("AudioTranscribe module")
|
|
|
|
class AudioTranscription:
|
|
model_name = "openai/whisper-large-v2"
|
|
|
|
filepath: str
|
|
waveform: torch.Tensor
|
|
sampling_rate: int
|
|
|
|
chunks: list = []
|
|
chunk_size: int
|
|
device = "cuda"
|
|
processor: WhisperProcessor
|
|
model: WhisperForConditionalGeneration
|
|
|
|
language = "ru"
|
|
|
|
all_transcription: list = []
|
|
|
|
def __init__(
|
|
self,
|
|
filepath: str,
|
|
language = "ru",
|
|
device = "cuda",
|
|
model_name = "openai/whisper-large-v2"
|
|
) -> None:
|
|
self.filepath = filepath
|
|
self.language = language
|
|
self.device = device
|
|
self.model_name = model_name
|
|
self.chunks: list = []
|
|
try:
|
|
logger.info("Loading model WhisperProcessor...")
|
|
self.processor = WhisperProcessor.from_pretrained(self.model_name)
|
|
|
|
self.model = WhisperForConditionalGeneration.from_pretrained(
|
|
self.model_name,
|
|
torch_dtype=torch.float16,
|
|
device_map="auto"
|
|
).to(self.device)
|
|
|
|
logger.info("Model loaded.")
|
|
|
|
self.waveform, self.sampling_rate = torchaudio.load(filepath, format="mp3", backend="ffmpeg")
|
|
logger.info(f"Successfully loaded file {filepath}.")
|
|
|
|
except Exception as e:
|
|
logger.error(f"Unable to load file {self.filepath}: {e}")
|
|
raise
|
|
|
|
def resample(self) -> None:
|
|
self.waveform = torchaudio.functional.resample(self.waveform, self.sampling_rate, 16000)
|
|
|
|
def to_mono(self):
|
|
if self.waveform.shape[0] > 1:
|
|
self.waveform = self.waveform.mean(dim=0, keepdim=True)
|
|
self.waveform = self.waveform.squeeze(0)
|
|
|
|
def split_to_chunks(self, chunk_length_s: int = 30) -> None:
|
|
logger.info(f"Splitting audio on chunks...")
|
|
|
|
self.chunk_size = chunk_length_s * 16000 # 16kHz after resampling
|
|
total_samples = self.waveform.shape[0]
|
|
chunks_count = (total_samples + self.chunk_size - 1) // self.chunk_size
|
|
|
|
logger.info(f"File length - {total_samples / 16000:.1f} seconds, splitting on {chunks_count} chunks by {chunk_length_s} seconds per chunk.")
|
|
|
|
self.chunks = []
|
|
for idx in range(chunks_count):
|
|
start = idx * self.chunk_size
|
|
end = min((idx + 1) * self.chunk_size, total_samples)
|
|
chunk = self.waveform[start:end].cpu().numpy().astype("float32")
|
|
self.chunks.append(chunk)
|
|
|
|
def process_chunk(
|
|
self,
|
|
chunk
|
|
) -> str:
|
|
inputs = self.processor(chunk, sampling_rate=16000, return_tensors="pt")
|
|
input_features = inputs.input_features.to(self.device).to(torch.float16)
|
|
|
|
with torch.no_grad():
|
|
predicted_ids = self.model.generate(
|
|
input_features,
|
|
language=self.language,
|
|
task="transcribe",
|
|
temperature=0.0
|
|
)
|
|
|
|
text = self.processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
|
|
|
|
return text
|
|
|
|
def process_all_chunks(self, batch_size: int = 16) -> None:
|
|
start_time = time.time()
|
|
try:
|
|
self.all_transcription = []
|
|
|
|
for i in tqdm(range(math.ceil(len(self.chunks) / batch_size))):
|
|
# TODO: rewrite batching as a separate function
|
|
batch = self.chunks[i*batch_size:(i+1)*batch_size]
|
|
|
|
inputs = self.processor(
|
|
batch,
|
|
sampling_rate=16000,
|
|
return_tensors="pt",
|
|
padding=True
|
|
)
|
|
|
|
input_features = inputs.input_features.to(self.device).to(torch.float16)
|
|
|
|
with torch.no_grad():
|
|
predicted_ids = self.model.generate(
|
|
input_features,
|
|
language=self.language,
|
|
task="transcribe",
|
|
temperature=0.0
|
|
)
|
|
|
|
texts = self.processor.batch_decode(predicted_ids, skip_special_tokens=True)
|
|
self.all_transcription.extend(texts)
|
|
|
|
end_time = time.time()
|
|
logger.info(f"Transcription completed in {end_time - start_time:.2f} seconds")
|
|
|
|
except Exception as e:
|
|
logger.error(f"Errors occured while processing chunks: {e}")
|
|
|
|
|
|
def transcribe_audio(self) -> str:
|
|
self.resample()
|
|
self.to_mono()
|
|
self.split_to_chunks()
|
|
self.process_all_chunks()
|
|
return " ".join(self.all_transcription) |