code error: AssertionError: assert T > 0

[hjc3613]

Backend impacted

The PyTorch implementation

Operating system

Linux

Hardware

CPU

Description

when I debug the code bellow copied from the readme file(moshi/README.md):

wave, sample_rate = torch.randn(1, 1, 24000 * 10), mimi.sample_rate
with torch.no_grad():
    codes = mimi.encode(wav)  # [B, K = 8, T]
    decoded = mimi.decode(codes)
    # save_as_mp3(decoded.numpy().squeeze(), sample_rate, audio_path.replace('.mp3', '_decoded.wav'))
    # Supports streaming too.
    frame_size = int(mimi.sample_rate / mimi.frame_rate)
    all_codes = []
    with mimi.streaming(batch_size=1):
        for offset in range(0, wav.shape[-1], frame_size):
            frame = wav[:, :, offset: offset + frame_size]
            codes = mimi.encode(frame)
            assert codes.shape[-1] == 1, codes.shape
            all_codes.append(codes)

I got the error:

what is the reason? how to solve it ?
I guess it's because of the last frame's shape is not the same as which before it. the last frame shape is : [1,1,363], others is [1,1,1920]

Extra information

I only cloned the repo, download the moshi weights, and run the test code from readme.md, I do not modify anything

Environment

Fill in the following information on your system.

• Operating system version:
  

If the backend impacted is PyTorch:

• Python version:3.12
• PyTorch version:

• Only runing in CPU