Anar Z. Yusifov retweetledi
Anar Z. Yusifov
587 posts

Anar Z. Yusifov
@anriumri
Math, HPC, SciFi, Soft Computing
Baku Katılım Mart 2010
226 Takip Edilen97 Takipçiler

@reidatcheson In terms of cutting a graph, I would use heuristic like "filling" the edges with some filler, then would apply temperature and shortly after that pressure. Iterative solver would eventually frac a stracture along the weak edges, but temperature would glue the strong ones together
English

@reidatcheson Have you tested it on diverse enough matrix structures? I never touched this subject (consciously), so I can't really reflect.
English

@lfortranorg I think you should focus on transpilation of the code to other languages and also refactoring through ASR of old codes into new codes...
English


@AnkaChen1 How does this compare to @taichi_lang in terms of performance and quality?
English

#SIGGRAPH2024 We present Vertex Block Descent (VBD), the NEXT simulator.:
nearly zero-compromised, fast🚀🚀, massively parallel💻💻, convergent📉📉, and unconditionally stable💎💎.
Project Page: ankachan.github.io/Projects/Verte…
Code (coming soon): github.com/AnkaChan/Gaia
English

@mrocklin Tell me your age without telling me your age :)) A bit one-sided and too specific, but yes - I agree.
English

Oof, just got this from Twitter.
Y'all are great and all, but it's time for a change 🙂.
#MyXAnniversary

English

@reidatcheson It's vaguely related, stumbled upon this while working on symbolically constructing matrices... "Twelve Proofs of the Markov Inequality
Aleksei Shadrin".
English

Screw Dask, I'm all in on Apache Spark now.
Coiled now supports Spark clusters
docs.coiled.io/user_guide/spa…
Read more to find out about this excellent feature
1/n
English

@historyinmemes Every PhD. owner did the same, and most of them never used that study ever again, just like Cooper.
English

@danbgoldman @docmilanfar I love this version, the most correct representation I've read. Reminds me on one of the articles about style transfer (can't find now 😞). However, it's just one interpretation/implementation of diffusion model. @docmilanfar has similar, more probabilistic interpretation.
English

@docmilanfar I found this derivation much easier to digest than others: ggx-research.github.io/publication/20…
(At least, for someone like me with a graphics background.)
English

@kolibril13 Just use @Panel_org - well maintained, provides a rich ecosystem and more deployment options, and integrates with third-party than anything else.
English

Recently read (audio book) The Ants by Boris Vian... About war, about its casual ugly side... Impressed is not the right word... appalled, shocked, stun, shattered by truth... #powerofwords #Literature #StopTheWar
GIF
English

@vchuravy Pixi + conda will do a trick: github.com/prefix-dev/pix…
English


@science_dot MPI is good for speed (productivity and performance) in HPC. However, I'm waiting for the moment when I can add and remove ranks (by allocating and deallocating machines) dynamically without restarting my application from the checkpoint. Dynamically scaling in and out for short.
English

I'll be sharing preliminary results on large data dataframe benchmarks in about an hour. It should be fun.
(I'll do something more serious/clean later as well. Maybe for PyData NYC?)
Dask@dask_dev
Join us for Dask demo day! There will be talks on Fondant, @RAPIDSai, Arraylake from @EarthmoverHQ, and TPC-H benchmarks for Spark, Dask, Polars, and DuckDB. 🗓️ Oct. 19 (today!) ⏰ 11am Eastern github.com/dask/community…
English

@OndrejCertik @wuoulf Sounds like it does something that was possible in Julia all this time. But it solves it for multiple languages, which is great!
English

Finally had a chance to play with Pixi (github.com/prefix-dev/pixi) from
@wuoulf & team, and it looks solid! I couldn't have done it better. It fixes an old issue I reported over 3 years ago at github.com/mamba-org/mamb…, but in a much better way than I had envisioned. Great job!!
English










