Gradio only using cpu

[Voltage71]

Checks

• This template is only for usage issues encountered.
• I have thoroughly reviewed the project documentation but couldn't find information to solve my problem.
• I have searched for existing issues, including closed ones, and couldn't find a solution.
• I confirm that I am using English to submit this report in order to facilitate communication.

Environment Details

OS: Windows 11
CPU: Ryzen 7 5800x
GPU: RX 6950 XT
Torch: 2.5.1
rocm6.2

When I try to synthesize a text, it shows 1 step and it takes up to 10 minutes to do one sentence. My CPU utilization jumps between 60% - 70% while my GPU stays at 2%.

I have tried this set up on my laptop that has Windows 11 and a GTX 1650 MaxQ and that also uses 100% of the CPU instead of the GPU.

What am I doing wrong? I pasted the notes it said when starting. Not sure what/where to change it.
P.S. I am pretty new at all of this.

Steps to Reproduce

I created a new conda environment and installed torch with pip install torch==2.5.1+rocm6.2 torchaudio==2.5.1+rocm6.2 --extra-index-url https://download.pytorch.org/whl/rocm6.2

I then did the pip install git+https://github.com/SWivid/F5-TTS.git

And I'm using the f5-tts_infer-gradio command after activating conda. When I hit Synthesize, I get the following message before it starts

"FutureWarning: The input name inputs is deprecated. Please make sure to use input_features instead.
warnings.warn(
You have passed task=transcribe, but also have set forced_decoder_ids to [[1, None], [2, 50360]] which creates a conflict. forced_decoder_ids will be ignored in favor of task=transcribe.
Passing a tuple of past_key_values is deprecated and will be removed in Transformers v4.43.0. You should pass an instance of EncoderDecoderCache instead, e.g. past_key_values=EncoderDecoderCache.from_legacy_cache(past_key_values).
The attention mask is not set and cannot be inferred from input because pad token is same as eos token. As a consequence, you may observe unexpected behavior. Please pass your input's attention_mask to obtain reliable results."

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

[SWivid]

I created a new conda environment and installed torch with pip install torch==2.5.1+rocm6.2 torchaudio==2.5.1+rocm6.2 --extra-index-url https://download.pytorch.org/whl/rocm6.2

#671 try linux or wsl

or use onnx version for windows

[Voltage71]

I created a new conda environment and installed torch with pip install torch==2.5.1+rocm6.2 torchaudio==2.5.1+rocm6.2 --extra-index-url https://download.pytorch.org/whl/rocm6.2

#671 try linux or wsl

or use onnx version for windows

Do you think this would work on a Linux VM?

[SWivid]

Do you think this would work on a Linux VM?

the link mentioned above said worked

[Voltage71]

Do you think this would work on a Linux VM?

the link mentioned above said worked

I borrowed a RTX 3060 12GB from a friend. I uninstalled all the AMD GPU drivers and installed the latest Nvidia drivers. I also installed cuda 11.8. I verified that cuda is installed by doing nvcc --version.

I then did

pip install torch==2.3.0+cu118 torchaudio==2.3.0+cu118 --extra-index-url https://download.pytorch.org/whl/cu118

And that installed successfully. I started up F5 and loaded in the sample. I tried generating text and the output is

Device set to use cpu C:\Users\User\.conda\envs\f5-tts\lib\site-packages\transformers\models\whisper\generation_whisper.py:573: FutureWarning: The input name inputsis deprecated. Please make sure to useinput_featuresinstead. warnings.warn( You have passed task=transcribe, but also have setforced_decoder_idsto [[1, None], [2, 50360]] which creates a conflict.forced_decoder_idswill be ignored in favor of task=transcribe. Passing a tuple ofpast_key_valuesis deprecated and will be removed in Transformers v4.43.0. You should pass an instance ofEncoderDecoderCacheinstead, e.g.past_key_values=EncoderDecoderCache.from_legacy_cache(past_key_values). The attention mask is not set and cannot be inferred from input because pad token is same as eos token. As a consequence, you may observe unexpected behavior. Please pass your input's attention_mask to obtain reliable results.

My cpu usage is 60% while my GPU usage is at 2%.
What am I missing?

[Voltage71]

Got it working!

The only thing I did was installed it again using Pinokio