cli.py It took too much time.

[fallbernana123456]

python CogVideoX/cli.py --lora_path TransPixar/cogvideox_rgba_lora.safetensors --prompt "A transparent glass of water with ice cubes gently swirling"

1/50 [00:11<2:58:21, 243.22s/it]
Is there something I set wrong?

[wileewang]

Could you provide more information about your GPU device? If you have limited resources, you can follow the Hugging Face demo settings to reduce the number of frames and inference timesteps. The generation process should take approximately 3–5 minutes.

[korober]

I have very similar results (2 / 25 ~ 45 min):

my setup: laptop 4090 16GB, Intel 14th + 64RAM

[jxwr]

I got 'UserWarning: Torch was not compiled with flash attention' . Is the slow speed related to this? My GPU is a 4090, 24GB.