Very poor performance on Windows

[amasu]

Hello,

First I would like to thank the creators and the contributors for this huge work, and for sharing the fruits :)

I am experimenting with genmo on Windows (Pycharm + a virtual environment).
Installed using pip ( Python 3.11) as per the README, then starting gradio UI with the same arguments as in the README.

I am am experiencing a surprisingly low performance: The estimated time to generate a video with the default prompt (31 frames, 480x848 , CFGscale=3, numInferences=100) is more than 12 hours at ~500s/it .

I am running pytorch + cuda 12.4 on an RTX 4090.

I don't know what to expect and I am mostly fishing for information here: Are these ETA and ratio normal ? Is there something I need to do to get a decent performance? What would be a decent performance?

Thank you for any feedback.

My Run log:

`INFO: Could not find files for the given pattern(s).

• Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Launching with 1 GPUs. If you want to force single GPU mode use CUDA_VISIBLE_DEVICES=0.
Attention mode: sdpa
Timing load_text_encoder
Timing load_dit
Timing load_vae
Stage Time(s) Percent
load_text_encoder 0.69 6.75%
load_dit 8.35 82.11%
load_vae 1.13 11.14%
Max memory reserved: 0.00 GB
moving model from cpu -> cuda:0
moving model from cuda:0 -> cpu
Max memory reserved: 17.85 GB
moving model from cpu -> cuda:0
Sampling: 19%|█▉ | 19/100 [2:32:00<11:23:06, 506.00s/it]
`