Michael Jungo

The hashes are `uint`s (represented as `number` for `uint32` and `bigint` for `uint64` in JS) and at no point are they stored as anything other than `uint`. I am assuming...

I can confirm this, Node seems to be significantly faster than Chrome and Firefox (didn't test Safari, as I'm on Linux). The timings have been run in incognito mode and...

xxh128 and also SIMD for xxh64 would certainly be interesting. Unfortunately, Safari still doesn't support WASM's SIMD, and I think that all major browsers should be supported without any feature...

Yes, absolutely. But I personally don't really have the time right now to look into the WebAssembly SIMD and to port xxh3 implementation to WebAssembly. I'm of course open for...

The padding is only used to be able to batch different length sequences and it is very easy for the model to learn that once it has seen all characters,...

I will not change it to use CTC loss as I will no longer modify anything on this model, since it was from a paper that I found interesting and...

That was the very first attempt to make it learn properly and at that time I didn't use a train/validation split, but used all samples for training. Since it is...

That is just not trained long enough. The accuracy on the training set should reach at least 95%, and if I remember correctly, the validation accuracy would be somewhere over...

You probably need to adjust the learning rate, including the options of the learning rate decay, as by default it reduces the learning rate by a factor of 10 every...

There is no implementation to get the LaTeX string from any image, because this project was used for research purposes and only the comparison to the test datasets was relevant....