|
--- |
|
title: README |
|
emoji: 🚀 |
|
colorFrom: green |
|
colorTo: indigo |
|
sdk: static |
|
pinned: false |
|
short_description: Empower AI inference |
|
--- |
|
|
|
<img width="35%" src="./kalray_logo.png"></a></br> |
|
|
|
Kalray enables AI innovators to build novel AI applications, maximizing your compute processing with MPPA Coolidge 2. Our |
|
compute Acceleration Cards offer a very complementary architecture to GPUs, allowing for the processing a large number |
|
of different operations in parallel in an asynchronous way. Details can be found here: |
|
* [Processor white paper](https://www.kalrayinc.com/resource/a6-mppa-coolidge-processor-white-paper/) |
|
* [Computation cards](https://www.kalrayinc.com/products/dpu-processors/#turbocard4) |
|
* [ML & computer vision](https://www.kalrayinc.com/solutions/#computer-vision) |
|
* [SDK description] |
|
|
|
You should find on this page several of the models that Kalray's SDK support. As part of ACE (AccessCore SDK), |
|
KaNN (Kalray Neural Network) is dedicated to optimize inference on the Kalray's processor (MPPA) on the following scheme: |
|
1. Design and/or import your Neural Networks from ONNX or TensorFlow (PyTorch is supported using the ONNX bridge), |
|
2. Build an intermediate represention of the NN in order to be executed on the MPPA, |
|
3. Run and exploit predictions from the device |
|
|
|
Find out on our github page the possibility to deploy and power your AI solutions over the Kalray's processor at: |
|
* [KANN-MODELS-ZOO](https://github.com/kalray/kann-models-zoo) |
|
* [WIKI](https://github.com/kalray/kann-models-zoo/blob/main/WIKI.md) |
|
|
|
Kalray 😎 |
