Dilshod Tadjibaev

I think natively we should stick to the file format we have currently, primarily because it's difficult to maintain many formats. We already support JSON (for visual debugging), Bincode (for...

@nathanielsimard it would be easy to do now that PyTorchFileRecorder works nicely. We just need to do the same things but using safetensors reader which is available from Candle.

Thanks for filing this. This is helpful as we prioritize ONNX ops. If you have a direct link to the ONNX file, can you also link this?

Updating Expand to Import, since we just added this op. I need to update the docs.

Submitted a PR to fix the supported OPs document: https://github.com/tracel-ai/burn/pull/1547

I think it might be faster to implement the model manually in Burn and load the pth weights file, which we now support. You can check out an existing model...

@laggui, if we decide to work on this, I am more inclined to adding ONNX OPs. It will be biggest bang for the buck instead of spending time to come...

Regarding Tile Op. We need to rename our current repeat op to repeat_dim and implement a proper repeat for all dimensions at once.

Resolving this ticket will resolve https://github.com/tracel-ai/burn/issues/1560 as well.

CC @nathanielsimard , @laggui , @louisfd