cestpasphoto

Adding my opinion on this (old) thread: I implemented the formula proposed above `pi -= (1-valids)*1000` I also implemented [this paper](https://arxiv.org/pdf/2006.14171.pdf) (check also the related github repo) with `pi =...

Wouldn't it be better to set the target of dummy points to what regressor would expect? I mean register_dummy() should use `optimizer._gp.predict(params)` as target.

Please use the following file: https://github.com/cestpasphoto/alpha-zero-general/blob/master/splendor/example_onnx_file.onnx To answer the auto-label, **CUDA is NOT involved** here, it happens with CPU-only environment.

I've tried different version of torch and onnxruntime, and it looks like export of some networks (including splendor one) aren't compatible with onnxruntime>=1.17.0, whatever the export settings are. That is...

I also raised https://github.com/microsoft/onnxruntime/issues/20641

Hi @lumi-a , I have seen your answer on ticket above: there isn't a proper fix but a workaround. I've added it in the code, which should now support all...

Correct. When using their snippet, I could convert the current ONNX to latest opset and onnxruntime can load such updated file. The problem is that pytorch can export into ONNX...

Unfortunately I won't be able to contribute in the near future: I'm a new dad ... with twins. Wish me luck! Tip: you may need to update the NN architecture...