Saleem DjimaandVáclav Volhejn
4ec89acf95
This pull request addresses issue #38 **Changes included:** - Added a `max_tokens` parameter to the typer cli for the `generate` command, allowing user to set a default number of tokens per chunk. - Introduced a default variable `MAX_TOKEN_PER_CHUNK` to centralize the default value. - Fixed a tokenization issue by inserting a comma, allowing the model to correctly read each chunk as intended rather than treating one chunk as a single sentence. **Future improvement suggestion :** - Consider enhancing the `split_into_best_sentences` function using regex to split text more accurately without modifying it, which would improve clarity and maintainability. --------- Co-authored-by: Václav Volhejn <8401624+vvolhejn@users.noreply.github.com>
29 lines
947 B
Python
29 lines
947 B
Python
"""Various utilities for audio conversion (pcm format, sample rate and channels),
|
|
and volume normalization."""
|
|
|
|
import torch
|
|
from scipy.signal import resample_poly
|
|
|
|
|
|
def convert_audio(
|
|
wav: torch.Tensor, from_rate: int | float, to_rate: int | float, to_channels: int
|
|
) -> torch.Tensor:
|
|
"""Convert audio to new sample rate and number of audio channels."""
|
|
if from_rate != to_rate:
|
|
# Convert to numpy for scipy resampling
|
|
wav_np = wav.detach().cpu().numpy()
|
|
|
|
# Calculate resampling parameters
|
|
gcd = int(torch.gcd(torch.tensor(from_rate), torch.tensor(to_rate)).item())
|
|
up = int(to_rate // gcd)
|
|
down = int(from_rate // gcd)
|
|
|
|
# Resample using scipy
|
|
resampled_np = resample_poly(wav_np, up, down, axis=-1)
|
|
|
|
# Convert back to torch tensor
|
|
wav = torch.from_numpy(resampled_np).to(wav.device).to(wav.dtype)
|
|
|
|
assert wav.shape[-2] == to_channels
|
|
return wav
|