https://github.com/patil-suraj/diffusers-dynamo-tests

https://github.com/patil-suraj/diffusers-dynamo-tests

Last synced: 1 day ago
JSON representation

• Host: GitHub
• URL: https://github.com/patil-suraj/diffusers-dynamo-tests
• Owner: patil-suraj
• Created: 2023-01-24T07:57:36.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2023-01-24T10:32:05.000Z (almost 2 years ago)
• Last Synced: 2024-12-15T12:41:51.914Z (26 days ago)
• Language: Python
• Size: 14.6 KB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
Repo to test PyTorch 2.0 for inference optimization on CPU and GPU

```pip install -r requirements.txt```

## Experiments

This folder contains quick reproducers to observe different behaviors. The base scripts give a benchmark without any optimization. Then we can observe what happens when optimizing different things. This table regroups the total time for 4 steps (all executed on an `Intel(R) Xeon(R) CPU @ 2.20GHz`):

We measure total time here since in diffusion inference we usually to multiple forward passes of the same model.

*For the optimized model the total time is measured after a warmup step.*

Batch size 4, steps 4:

| Script | Toal time |

|:--|:-:|

| base_cpu | 19.36s |

| optimize_cpu | 18.69s |

Batch size 8, steps 4:

| Script | Toal time |

|:--|:-:|

| base_cpu | 31.36s |

| optimize_cpu | 38.69s |