-
-
Notifications
You must be signed in to change notification settings - Fork 72
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.
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
Make a onnx-go compatible with tiny Go #102
Comments
With this piece of code : package main
import (
"fmt"
"github.com/owulveryck/onnx-go/backend/x/gorgonnx"
)
func main() {
backend := gorgonnx.NewGraph()
fmt.Println(backend)
} We have the following error :
I will try to stub the method Comparable in tinygo. |
Just stub the method and don't work.
Then I remove
|
Also there is a dependencies on uuid (which is linked to sql/drivers) and tinygo is not compatible to this module right now. |
With gorgonia (based on the example) :
Will stub this method tomorrow. |
UUID is linked with sql/driver?
I think we can get rid of this. We need to find a way to generate names and
avoid conflicts in our namespace.
|
Just found this issue from Twitter! Reflect support in TinyGo is still very minimal and doesn't support the JSON package yet. I hope that I can improve reflect support in the coming weeks so it becomes usable for serialization/deserialization. |
Yes* (and in general uuid is linked to sql^^), what are the restrictions of generations ? *:https://github.com/google/uuid/blob/master/sql.go#L8 |
PR #103 removes the need for the |
What would be the primary use case, WebAssembly or embedded devices? For embedded devices, this may be of interest: |
Webassembly is a target. It will allow the developers to build new kind of "web applications" involving neural networks easily. For example, let's consider a website that would do sentiment analysis on a picture. With the neural net coded into the webpage via WASM, no need to upload the photo anywhere (in term of ethics and compliance this would be, IMHO, a step forward). But, TBH, that would also be very interesting to be able to run a neural net smoothly on an embedded device. This would give the possibility for the developer to add "smart capacities" to their objects. Thank you for the link, I did not know about that; From what I see, this could be something to implement on the same level as the Cuda binding directly into the Gorgonia engine. I will have a closer look. I hope this answers your questions, otherwise, do not hesitate to reply here (or to reach me directly, I am @owulveryck on gopher slack) |
Might be worth trying again. |
This issue will track the work done to make the package compatible with tiny-go.
The computation backend is, by now outside of this experiment.
Once the package works with tiny go, we may write a very simple backend to use it for example in webassembly.
The text was updated successfully, but these errors were encountered: