Timothy Mak

Have you tried uninstalling lassosum and re-installing?

You mean the p-value is less than the 2.2e-308 (i.e. `double.xmin`)? If so, I think a better idea is to replace these with `double.xmin`, since low p-values are supposed to...

Wow, thanks very much. I think I understand now. That's very helpful of you.

Have you tried the recommended actions, e.g. increasing `mem.limit`? lassosum assumes your memory is limited (c.f. [here](https://github.com/tshmak/lassosum/blob/0e44b530c77e862f1b88f9e1fad88fdab44e7827/R/lassosum.R#L49)). If you don't want to increase, then try the other options (split into...

I can't replicate the difference using the toy example I provided in README.md. Also, it is supposed to be the simple Pearson's correlation and not R2 that is being calculated.

I guess I had a misconception about huggingface tokenizers. It seems in general, these tokenizers are themselves neural networks and their outputs may be somewhat unpredictable.

But how does it determine where to insert the '|'? The output here, ``` ɡ ei2 j a4 n [UNK] ɡ eɜ p anɡ4 j au5 t aaɜ n [UNK]...

I'm also experiencing this issue. I'm applying Aero to a bunch of non-English data, and most of my wavfiles had this issue. Appreciate your effort into sharing your trained model....

Actually, I found that this problem affects the 12-48_512_128 checkpoint as well. So the model itself is buggy.

I don't want to sound unappreciative of you or the author for sharing your codes and the models to us *for free* and with a permissive license. I do agree...