Skip to content

fixed the _inference_sample method to use the correct configuration p… #30

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
AdrLfv wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fixed the _inference_sample method to use the correct configuration p… #30

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 28 additions & 7 deletions models/SepReformer_Base_WSJ0/engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -150,13 +150,32 @@ def _test(self, dataloader, wav_dir=None):

@logger_wraps()
def _inference_sample(self, sample):
self.model.eval()
self.fs = self.config["dataset"]["sampling_rate"]
mixture, _ = librosa.load(sample,sr=self.fs)
mixture = torch.tensor(mixture, dtype=torch.float32)[None]
self.stride = self.config["model"]["module_audio_enc"]["stride"]
logger.info(f'Inference with {sample}')
self.fs = self.config['dataset']['sampling_rate']
self.stride = self.config['model']['module_audio_enc']['stride']

# Create output directory
if self.out_wav_dir is None:
output_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "output")
else:
output_dir = self.out_wav_dir

os.makedirs(output_dir, exist_ok=True)
sample_basename = os.path.basename(sample)
output_base = os.path.join(output_dir, os.path.splitext(sample_basename)[0])

logger.info(f'Output files will be saved to {output_dir}')

mixture, fs = sf.read(sample)
if fs != self.fs:
logger.warning(f'Resample from {fs} to {self.fs}')
mixture = librosa.resample(mixture, orig_sr=fs, target_sr=self.fs)

mixture = torch.tensor(mixture, dtype=torch.float32).unsqueeze(0).to(self.device)

remains = mixture.shape[-1] % self.stride
if remains != 0:
logger.info(f"Pad {remains} samples at the end with zeros.")
padding = self.stride - remains
mixture_padded = torch.nn.functional.pad(mixture, (0, padding), "constant", 0)
else:
Expand All @@ -166,10 +185,12 @@ def _inference_sample(self, sample):
nnet_input = mixture_padded.to(self.device)
estim_src, _ = torch.nn.parallel.data_parallel(self.model, nnet_input, device_ids=self.gpuid)
mixture = torch.squeeze(mixture).cpu().numpy()
sf.write(sample[:-4]+'_in.wav', 0.9*mixture/max(abs(mixture)), self.fs)
sf.write(f'{output_base}_in.wav', 0.9*mixture/max(abs(mixture)), self.fs)
logger.info(f'Saved input file: {output_base}_in.wav')
for i in range(self.config['model']['num_spks']):
src = torch.squeeze(estim_src[i][...,:mixture.shape[-1]]).cpu().data.numpy()
sf.write(sample[:-4]+'_out_'+str(i)+'.wav', 0.9*src/max(abs(src)), self.fs)
sf.write(f'{output_base}_out_{i}.wav', 0.9*src/max(abs(src)), self.fs)
logger.info(f'Saved output file {i}: {output_base}_out_{i}.wav')


@logger_wraps()
Expand Down