[Bug] xtts_v2, AttributeError: 'TTS' object has no attribute 'speakers'

[chigkim]

Describe the bug

If I run tts.speakers after loading xtts_v2, it throws an error: 'TTS' object has no attribute 'speakers'.

To Reproduce

import torch
from TTS.api import TTS
device = "cuda" if torch.cuda.is_available() else "cpu"
model = "tts_models/multilingual/multi-dataset/xtts_v2"
tts = TTS(model).to(device)
tts.speakers

Expected behavior

It should list all the speaker names.

Logs

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>.venv\Lib\site-packages\torch\nn\modules\module.py", line 1709, in __getattr__
    raise AttributeError(f"'{type(self).__name__}' object has no attribute '{name}'")
AttributeError: 'TTS' object has no attribute 'speakers'

Environment

{
    "CUDA": {
        "GPU": [],
        "available": false,
        "version": null
    },
    "Packages": {
        "PyTorch_debug": false,
        "PyTorch_version": "2.3.0+cpu",
        "TTS": "0.22.0",
        "numpy": "1.26.4"
    },
    "System": {
        "OS": "Windows",
        "architecture": [
            "64bit",
            "WindowsPE"
        ],
        "processor": "Intel64 Family 6 Model 94 Stepping 3, GenuineIntel",
        "python": "3.11.9",
        "version": "10.0.19045"
    }
}

Additional context

Other multi speaker models like vctk/vits work fine using the same method.

[eginhard]

This is fixed by eginhard#8 in our fork, available via pip install coqui-tts

[chigkim]

Thanks@eginhard! Had no idea this repo was abandoned.

[jamesob]

Thanks @eginhard. This repo should probably be marked as read-only/archive by its owners.

[Asrithapenmetsa]

Hi, @chigkim
Have you finetuned the vits model with your own dataset?
Because I am encountering this error AttributeError: 'TTS' object has no attribute 'get_model' while I am tryng to finetune the model?

[chigkim]

It's a bug, and it's fixed in another fork.
https://github.com/idiap/coqui-ai-TTS

[eginhard]

@Asrithapenmetsa Placing code between triple backticks would make it easier to read... Could you also share the full error trace, not just the last line?

[eginhard]

Your error message

    model = TTS('tts_models/en/ljspeech/vits').to(device)
NameError: name 'device' is not defined

tells you that the variable device is not defined. So add a line that defines it before, e.g.

device = "cuda" if torch.cuda.is_available() else "cpu"

[Asrithapenmetsa]

Thank you so much . It got fixed. Thanks

…

On Sun, 28 Jul 2024 at 01:41, Enno Hermann ***@***.***> wrote: Your error message model = TTS('tts_models/en/ljspeech/vits').to(device)NameError: name 'device' is not defined tells you that the variable device is not defined. So add a line that defines it before, e.g. device = "cuda" if torch.cuda.is_available() else "cpu" — Reply to this email directly, view it on GitHub <#3742 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BAJMUGOZLVVHMNKNANNWPXDZOP5HXAVCNFSM6AAAAABH2IMAYSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENJUGI2DGMRSGE> . You are receiving this because you were mentioned.Message ID: ***@***.***>