Closed samjgorman closed 2 years ago
Thanks for your feedback.
Large models consumes a lot of resources in browsers ( and also GPU resources if webgl backend ), which makes it unstable. For web scenario, usually 10MB~20MB is stable enough in my experience.
we are working on migrating ONNX.js to ONNX Runtime Web which offers enhanced user experience and improved performance. Please visit ONNX Runtime Web to get more details.
Appears that the large size of ResNet50_8.onnx (94mb) triggers a "Page Unresponsive" consistently when replicated locally on Chrome 92.0. What size models can onnxjs reliably serve on Chrome?
Edit: I quit Chrome, restarted and was able to successfully run the ResNet example. This may be a local issue, but will leave this issue open in case others have experienced other performance issues on Chrome.