enpasos

Results 7 comments of enpasos

In the example I set up for onnx export I am facing the following specific inheritance problems: ![grafik](https://user-images.githubusercontent.com/9622808/149083162-3a429440-ae87-47aa-8c0e-cc03c861015b.png) From the screenshot you see a particular problem with Conv2d with **package...

To achieve an ONNX export for my MuZero implementation https://enpasos.ai/muzero/How#onnx I only "opened" the four classes Conv2d, Linear, LayerNorm, LambdaBlock a little as mentioned above. What do you think about:...

@frankfliu Thank you for your expert look into the subject. Your converter approach gives a nice separation of concerns - much better than the inheritance approach I used. Looks promissing...

@frankfliu For me it would be perfect to switch to your approach. To be able to map an onnx join function to a djl join function in ParallelBlocks it would...

Two reasons why I closed the old issue and opened a new one (https://github.com/deepjavalibrary/djl/pull/2231): - I was reorganizing the branches of my fork - I reduced the content of the...

Hi @Artoriuz, I’m also interested in this functionality and couldn’t find a direct solution in the JAX ecosystem. So, I gave it a try and started working on a project...

[`Netron`](https://github.com/lutzroeder/netron) has just released version **8.2.6**, enhancing its support for **hierarchical function-based graph views** — and it genuinely feels a bit like discovering Google Maps for the first time. You...