HackerTrans
TopNewTrendsCommentsPastAskShowJobs

sdrg822

no profile record

comments

sdrg822
·anno scorso·discuss
“Model Control Plane” - quite early to already re-use a popular acronym
sdrg822
·anno scorso·discuss
Congrats on the launch!
sdrg822
·2 anni fa·discuss
+1 it wouldn’t be terribly useful if it were only caching the tokenizer output.
sdrg822
·2 anni fa·discuss
Where is the lie in this repo?
sdrg822
·2 anni fa·discuss
I think if you actually meet the people you'd realize they're pretty earnest and candid of these things' limitations, though it may not show in tweets and hackernews posts.

This repo, for instance, makes no claims of AGI. It just claims to help bootstrap a starting point for a software project: "it will not attempt to write the logic to fill in the nodes and edges."
sdrg822
·2 anni fa·discuss
But indexing *is* training. It's just not using end-to-end gradient descent.
sdrg822
·2 anni fa·discuss
Dang they use Transformer-XL from 2019 haha - didn't realize people still used that / XLNet-like architectures
sdrg822
·2 anni fa·discuss
It is not.
sdrg822
·2 anni fa·discuss
Congrats on the launch!