TypeError: Lm.warmup() missing 1 required positional argument: 'ct'

[hzeng-otterai]

Backend impacted

The MLX implementation

Operating system

Linux

Hardware

CPU

Description

Hi, I am running the following command on a Mac laptop and got the following error. Searched for it but can't find any related issues.

% python -m moshi_mlx.local_web
[Info] [SERVER] loading text tokenizer /Users/***/.cache/huggingface/hub/models--kyutai--moshiko-mlx-bf16/snapshots/a31ebad976783a3ea356d6beb8b8e826d2d45a68/tokenizer_spm_32k_3.model
[Info] [SERVER] loading weights /Users/***//.cache/huggingface/hub/models--kyutai--moshiko-mlx-bf16/snapshots/a31ebad976783a3ea356d6beb8b8e826d2d45a68/model.safetensors
[Info] [SERVER] weights loaded
Process Process-2:
Traceback (most recent call last):
  File "/Users/***//miniconda3/envs/moshi/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/Users/***//miniconda3/envs/moshi/lib/python3.12/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/Users/***//miniconda3/envs/moshi/lib/python3.12/site-packages/moshi_mlx/local_web.py", line 132, in model_server
    model.warmup()
TypeError: Lm.warmup() missing 1 required positional argument: 'ct'

Extra information

As above

Environment

Fill in the following information on your system.

• Operating system version: Mac OS 15.3

If the backend impacted is PyTorch:

• Python version:
• PyTorch version:
• CUDA version (run python -c 'import torch; print(torch.version.cuda)'):
• GPU model and memory:

If the backend is MLX:

• Mac model: Apple M1 Pro with 16GB memory, Mac OS 15.3

[hzeng-otterai]

Python version is 3.12.9.

[CuriousMonkey7]

same issue

[CuriousMonkey7]

@hzeng-otterai It turns out they forgot to set the default value of ct to None. This has been fixed in the code now.

moshi/moshi_mlx/moshi_mlx/models/lm.py

Lines 304 to 311 in fcc02ca

	def warmup(self, ct: ConditionTensor | None = None):
	text, audio = self.sample(
	mx.array([[self.cfg.text_out_vocab_size]]),
	[mx.array([[0]])] * self.cfg.other_codebooks,
	text_sampler=sampling.Sampler(),
	audio_sampler=sampling.Sampler(),
	ct=ct,
	)

You can uninstall the current version and use the following command to install the latest version:

pip install -U -e "git+https://[email protected]/kyutai-labs/moshi.git#egg=moshi_mlx&subdirectory=moshi_mlx"