out of Memory - with 8 GB VRAM RTX 2080 super

fudan-generative-vision / hallo2

Hallo2: Long-Duration and High-Resolution Audio-driven Portrait Image Animation

https://fudan-generative-vision.github.io/hallo2

MIT License

4.38k stars 625 forks source link

out of Memory - with 8 GB VRAM RTX 2080 super #16

Open sugatasanshiro opened 1 month ago

sugatasanshiro commented 1 month ago

cmd_OfMxMsnDwa

cmd_CNAzTnGcdA

cuijh26 commented 1 month ago

So sorry to hear that. Our model need about 13GB memory at least. We are training the lightweight version right now.

sugatasanshiro commented 1 month ago

oh good

AricGamma commented 1 month ago

duplicate of #15

MftFrank commented 4 weeks ago

I'm using a Tesla M60 graphics card, which has two GPU cores, each with 8GB of memory. When I run my program, I still encounter out-of-memory issues. How should I configure my setup to use both GPUs?

superlbr commented 2 weeks ago

export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True 3060 12GB work