@tensorflow/tfjs-backend-webgpu v4.19.0
Usage
This package adds a GPU accelerated WebGPU backend to TensorFlow.js. It currently supports the following models:
- BlazeFace
- BodyPix
- Face landmarks detection
- HandPose
- MobileNet
- PoseDetection
- Universal sentence encoder
- AutoML Image classification
- AutoML Object detection
- Speech commands
Google Chrome started to support WebGPU by default in M113 on May 2, 2023.
Importing the backend
Via NPM
// Import @tensorflow/tfjs or @tensorflow/tfjs-core
import * as tf from '@tensorflow/tfjs';
// Add the WebGPU backend to the global backend registry.
import '@tensorflow/tfjs-backend-webgpu';
// Set the backend to WebGPU and wait for the module to be ready.
tf.setBackend('webgpu').then(() => main());
Via a script tag
<!-- Import @tensorflow/tfjs or @tensorflow/tfjs-core -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs/dist/tf.min.js"> </script>
<!-- Add the WebGPU backend to the global backend registry -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs-backend-webgpu/dist/tf-backend-webgpu.js"></script>
<script>
// Set the backend to WebGPU and wait for the module to be ready
tf.setBackend('webgpu').then(() => main());
</script>
FAQ
When should I use the WebGPU backend?
The mission of WebGPU backend is to achieve the best performance among all approaches. However, this target can not be met overnight, but we are committed to supporting it with rapid and continuous performance improvement. Many exciting features, like FP16, DP4A, will be brought in soon.
How many ops have you implemented?
See register_all_kernels.ts
for an up-to-date list of supported ops. We love contributions. See the
contributing
document for more info.
Do you support training?
Maybe. There are still a decent number of ops that we are missing in WebGPU that are needed for gradient computation. At this point we are focused on making inference as fast as possible.
Do you work in node?
Yes. If you run into issues, please let us know.
How do I give feedback?
We'd love your feedback as we develop this backend! Please file an issue here.
Development
Building
yarn build
Testing
Currently the Canary channel of Chrome is used for testing of the WebGPU backend:
yarn test # --test_env=CHROME_CANARY_BIN=/path/to/chrome
10 days ago
16 days ago
23 days ago
29 days ago
3 months ago
4 months ago
4 months ago
4 months ago
5 months ago
5 months ago
5 months ago
10 months ago
6 months ago
7 months ago
6 months ago
8 months ago
7 months ago
6 months ago
9 months ago
11 months ago
11 months ago
1 year ago
12 months ago
1 year ago
1 year ago
1 year ago
2 years ago
1 year ago
2 years ago
2 years ago
2 years ago
2 years ago
2 years ago
3 years ago
3 years ago
3 years ago
3 years ago
4 years ago
5 years ago
5 years ago
5 years ago
5 years ago