Angus Turner

I'd like to help! I'm quite new to ML, but would like to have a stab at Levenberg-Marquardt optimisation (or similar). At a glance, the maths doesn't seem _too difficult_,...

Maybe this deserves a separate issues / feature request, but it was non-obvious to me as a first-time user (coming from Haiku) how to overwrite the default initialization scheme in...

That is a fair point! It is true that PyTorch doesn't take a `w_init`, but because stuff is mutable in PyTorch its very easy to just set `linear.bias = ...`...

Hi, thanks for getting back to me! (Feel like I’m opening a new question everyday at this point 😅). I used the no-op as an illustrative example and to diagnose...

Update: issue seems to be resolved when I disabled `tqdm`. There must be a way to fix though?

I am finding that following @tautomer 's suggestion I can indeed get spaces switching to work, BUT at the cost of some weird sleep issues (related to #143 I think?)....

Actually, while I'm here, can I also clarify the interpretation of the dimensions, referenced in the docs? ``` event_shape (N x M x 3), e.g. phi(i, j, op) Ops are...

As far as I can tell, the issue seems to be some kind of bot protection / obfuscation that has been added to website? Something called KPSDK?

After some more digging, it seems like if you trigger the bot detection then instead of receiving a JSON response (as expected), you will instead get a 429 error code,...

Hi, sorry I didn't get back to you sooner! I have not tried fitting the GMM directly with gradient ascent, although it should be easy enough to test. I remember...