Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update README #847

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Update README #847

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
3a4493d
Update README.md
monorimet Sep 24, 2024
f31bcc7
Update README.md
monorimet Sep 24, 2024
cfaf0c2
Update README.md
monorimet Sep 24, 2024
68bfcb7
Update README.md
monorimet Sep 24, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 13 additions & 12 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,8 @@
# SHARK Turbine
# SHARK-ModelDev

This repo is Nod-AI's integration repository for various model bringup
activities and CI. In 2023 and early 2024, it played a different role
This is the AMD SHARK team's integration repository that connects inference tasks, such as Stable Diffusion, from their various source libraries to the IREE/SHARK ML acceleration and deployment framework.

In 2023 and early 2024, it played a different role
by being the place where FX/Dynamo based torch-mlir and IREE toolsets
were developed, including:

Expand All @@ -11,20 +12,20 @@ were developed, including:
* [IREE Turbine](https://github.com/iree-org/iree-turbine)
* [Sharktank and Shortfin](https://github.com/nod-ai/sharktank)

As these have all found upstream homes, this repo is a bit bare. We will
continue to use it as a staging ground for things that don't have a
more defined spot and as a way to drive certain kinds of upstreaming
activities.


## Current Projects
As these have all found upstream homes, this repo is now responsible for:
- Exporting (via iree-turbine tooling) and compiling (via iree-compile) torch modules from various sources, mostly huggingface libraries
- Carrying alternative (maximally exported and compiled) implementations to complex inference tasks e.g. Stable Diffusion (1.5, 2.1, SDXL, SD3, Flux)

### turbine-models

The `turbine-models` project (under models/) contains ports and adaptations
of various (mostly HF) models that we use in various ways.

### CI
The only implementation in turbine-models that is currently in use is its stable diffusion exports and pipeline, which are slated for migration to https://github.com/nod-ai/SHARK-Platform for productionization -- it is to be temporarily preserved here for proof of concept, functionality, and performance. The current state of the turbine-models SD implementation is heavily based on Diffusers' StableDiffusionPipelines, with the harnessing and actual inference code rewritten to offload as much as possible from torch (cpu) to the IREE compiler and runtime.

In the near future, the remaining purpose of turbine-models is to maintain and validate the " library import -> nn.module -> iree-turbine (dynamo) -> iree-compile " export and compilation stack, for a few different key model classes or suites.
The model validation and benchmarking of compiled artifacts is under migration to [SHARK-TestSuite](https://github.com/nod-ai/SHARK-TestSuite)

Integration CI for a variety of projects is rooted in this repo.
### CI / Tracking

A number of model support tasks are tracked in this repo's issues, and its CI is designed to be the origin point for relevant MLIR/VMFB artifacts used further downstream in test/benchmark/regression suites.
Loading