zhangyiss/mlx-examples
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx-examples.git synced 2025-08-09 10:26:38 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Clarification in the stable diffusion readme

Browse Source
This commit is contained in:
Angelos Katharopoulos 2023-12-06 15:22:46 -08:00 committed by Angelos Katharopoulos
parent e488831e03
commit a1b4a475e1
1 changed files with 7 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
stable_diffusion/README.md
View File

@ -65,8 +65,9 @@ Performance
----------- -----------
The following table compares the performance of the UNet in stable diffusion. The following table compares the performance of the UNet in stable diffusion.
We report throughput in images per second for the provided `txt2image.py` We report throughput in images per second **processed by the UNet** for the
script and the `diffusers` library using the MPS PyTorch backend. provided `txt2image.py` script and the `diffusers` library using the MPS
PyTorch backend.
At the time of writing this comparison convolutions are still some of the least At the time of writing this comparison convolutions are still some of the least
optimized operations in MLX. Despite that, MLX still achieves **~40% higher optimized operations in MLX. Despite that, MLX still achieves **~40% higher
@ -93,3 +94,7 @@ The above experiments were made on an M2 Ultra with PyTorch version 2.1,
diffusers version 0.21.4 and transformers version 4.33.3. For the generation we diffusers version 0.21.4 and transformers version 4.33.3. For the generation we
used classifier free guidance which means that the above batch sizes result used classifier free guidance which means that the above batch sizes result
double the images processed by the UNet. double the images processed by the UNet.
Note that the above table means that it takes about 90 seconds to fully
generate 16 images with MLX and 50 diffusion steps with classifier free
guidance and about 120 for PyTorch.