Michael Voznesensky

[Do not land] [RFC] 1.375x speedup - Remove control flow from model, small hacks, enable TorchDynamo + TorchInductor

14

Obviously not meant to land, this RP is representative of what it would take to get dynamo working. test_me.py takes 4.4 seconds on main branch test_me.py takes 3.2 seconds in...

Guard on at::Tensor device index

14

Fixes #91777 cc @mlazos @soumith @yanboliang @penguinwu @anijain2305 @EikanWang @jgong5 @Guobing-Chen @chunyuan-w @XiaobingSuper @zhuhaozhe @blzheng @Xia-Weiwen @wenzhe-nrv @jiayisunx @desertfire

Remove temporary export skip hack

2

Fixes #ISSUE_NUMBER cc @mlazos @soumith @yanboliang @penguinwu @anijain2305 @EikanWang @jgong5 @Guobing-Chen @chunyuan-w @XiaobingSuper @zhuhaozhe @blzheng @Xia-Weiwen @wenzhe-nrv @jiayisunx @desertfire

Refactor dynamic dims api, stateless internals, higher level export API

5

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #96699 The purpose of this API is to execute a few large components of work: 1) Refactor all the internals of plumbing...

### Description I am writing a pallas kernel, where one of the lines: `frobenius_sq_norm = square_norm(w_tl).sum(keepdims=True)...` includes a sum() w/ keepdims. I don't expect sum() to work without keepdims, at...

Message not encrypted

13

So - my goal is, in the least number of steps get a message encrypted with libotr on iOS Currently, with OTRKit, I have a class as an OTR Delegate,...

Bob And Alice

3

A bit long winded, but I am at a roadblock. First of all, thank you so much for this library/project. Internet security, privacy, and end-to-end encryption has never been more...