4 ms·
Just tried glid-3 with a batch size of one and I'm getting 4781MiB. The latent diffusion model peaks at 8403MiB These are fp16 numbers though, you might need a
by Jack000 5y ago
Just tried glid-3 with a batch size of one and I'm getting 4781MiB. The latent diffusion model peaks at 8403MiB
These are fp16 numbers though, you might need a recent nvidia card to run it.
- rahimnathwani 5y agoI'll try them out. I have an RTX 2070, which apparently supports fp16. But it only has 8GB RAM. I used the instructions here to check: https://github.com/wang-xinyu/tensorrtx/blob/master/tutorials/check_fp16_int8_support.md https://github.com/wang-xinyu/tensorrtx/blob/master/tutorial...