Error on the random voice example

[vroger11]

Using the Random voice example from the readme I got this error:

Traceback (most recent call last):
File "/Users/vincent.roger/git/parler-tts-test/test.py", line 18, in
generation = model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/vincent.roger/git/parler-tts-test/.venv/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/Users/vincent.roger/git/parler-tts-test/.venv/lib/python3.12/site-packages/parler_tts/modeling_parler_tts.py", line 3425, in generate
model_kwargs["attention_mask"] = self._prepare_attention_mask_for_generation(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/vincent.roger/git/parler-tts-test/.venv/lib/python3.12/site-packages/transformers/generation/utils.py", line 585, in _prepare_attention_mask_for_generation
pad_token_id = generation_config._pad_token_tensor
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'Tensor' object has no attribute '_pad_token_tensor'

Here are the versions of the libraries I used:
python: 3.12
accelerate: 1.3.0
soundfile: 0.13.0
torch: 2.5.1
transformers: 4.48.1
parler-tts: 0.2.2 with commit version: d108732

[ptrxwsmitt]

got the same issue:
python: 3.12
accelerate: 1.3.0
soundfile: 0.13.0
torch: 2.5.1
transformers: 4.48.1
parler-tts -> did not check commit yet, but latest commit from github as documentation

UPDATE:
After reading examples from https://github.com/huggingface/parler-tts/blob/main/INFERENCE.md and some experimenting it seems that we actually need to set the attention_mask and prompt_attention_mask in the Random voice example.

The follwing example is for the prompt. Do the same for the description:

prompt_tokenized = tokenizer(
    text=prompt,
    return_tensors="pt",
)
prompt_input_ids = prompt_tokenized.input_ids.to(device)
prompt_attn_mask = prompt_tokenized.attention_mask.to(device)

That worked for me