Matthew Muckley
@mattmucklm.bsky.social
Research Engineer, Meta Fundamental AI Research (FAIR). ML for compression, computer vision, medicine.
https://mmuckley.github.io/
https://mmuckley.github.io/
Very excited to share V-JEPA 2! I've been working on the encoder pretraining pipeline and data curation for this model the last few months, and am excited for it to finally be out!
ai.meta.com/blog/v-jepa-...
ai.meta.com/blog/v-jepa-...
ai.meta.com
June 11, 2025 at 9:43 PM
Very excited to share V-JEPA 2! I've been working on the encoder pretraining pipeline and data curation for this model the last few months, and am excited for it to finally be out!
ai.meta.com/blog/v-jepa-...
ai.meta.com/blog/v-jepa-...
We just published "Qinco2: Vector Compression and Search with Improved Implicit Neural Codebooks" on arXiv, work led by our talented intern, Theophane Vallaeys.
Qinco2 achieves as much as 40-60% reduction error for vector compression, as well as better performance for approximate similarity search.
Qinco2 achieves as much as 40-60% reduction error for vector compression, as well as better performance for approximate similarity search.
Qinco2: Vector Compression and Search with Improved Implicit Neural Codebooks
Vector quantization is a fundamental technique for compression and large-scale nearest neighbor search. For high-accuracy operating points, multi-codebook quantization associates data vectors with one...
arxiv.org
January 7, 2025 at 2:46 PM
We just published "Qinco2: Vector Compression and Search with Improved Implicit Neural Codebooks" on arXiv, work led by our talented intern, Theophane Vallaeys.
Qinco2 achieves as much as 40-60% reduction error for vector compression, as well as better performance for approximate similarity search.
Qinco2 achieves as much as 40-60% reduction error for vector compression, as well as better performance for approximate similarity search.
One thing I've found in research is the constant counterbalancing between "prototype" code and "engineered" code.
Prototyped code is often a bit hacky, but gets the job done. But if you ever need to extend it, it can be quite a pain.
Engineered code usually has some overarching design philosophy
Prototyped code is often a bit hacky, but gets the job done. But if you ever need to extend it, it can be quite a pain.
Engineered code usually has some overarching design philosophy
December 13, 2024 at 3:26 PM
One thing I've found in research is the constant counterbalancing between "prototype" code and "engineered" code.
Prototyped code is often a bit hacky, but gets the job done. But if you ever need to extend it, it can be quite a pain.
Engineered code usually has some overarching design philosophy
Prototyped code is often a bit hacky, but gets the job done. But if you ever need to extend it, it can be quite a pain.
Engineered code usually has some overarching design philosophy
For MRI folks: we just rolled out a new release to torchkbnufft, first in a couple years.
The changes are for working with newer package versions. Things now work on numpy 2.0, and a few deprecations are fixed. Other than that, it's the same as before :). Get it with
`pip install torchkbnufft`
The changes are for working with newer package versions. Things now work on numpy 2.0, and a few deprecations are fixed. Other than that, it's the same as before :). Get it with
`pip install torchkbnufft`
Release v1.5.2 Fix required numpy version · mmuckley/torchkbnufft
What's Changed
Update required numpy by @mmuckley in #103
Full Changelog: v1.5.1...v1.5.2
github.com
December 4, 2024 at 3:26 PM
For MRI folks: we just rolled out a new release to torchkbnufft, first in a couple years.
The changes are for working with newer package versions. Things now work on numpy 2.0, and a few deprecations are fixed. Other than that, it's the same as before :). Get it with
`pip install torchkbnufft`
The changes are for working with newer package versions. Things now work on numpy 2.0, and a few deprecations are fixed. Other than that, it's the same as before :). Get it with
`pip install torchkbnufft`
Good thoughts, some of which I've learned from trial and error over the years.
The advice part, centering things on technical points, is also useful for academic publishing and the review process. It really helps defuse what tends to be an adversarial relationship with reviewers (or authors).
The advice part, centering things on technical points, is also useful for academic publishing and the review process. It really helps defuse what tends to be an adversarial relationship with reviewers (or authors).
2. Then there's this very inspirational post from @gaelvaroquaux.bsky.social about handling technical discussions gael-varoquaux.info/programming/...
My takeaways (i) There's no you nor I, there are only features, (ii) be techincal and actionable, (iii) don't write anything you could regret
My takeaways (i) There's no you nor I, there are only features, (ii) be techincal and actionable, (iii) don't write anything you could regret
Technical discussions are hard; a few tips -- Gaël Varoquaux: computer / data / health science
Gaël Varoquaux, computer / data / health science
gael-varoquaux.info
November 23, 2024 at 3:46 PM
Good thoughts, some of which I've learned from trial and error over the years.
The advice part, centering things on technical points, is also useful for academic publishing and the review process. It really helps defuse what tends to be an adversarial relationship with reviewers (or authors).
The advice part, centering things on technical points, is also useful for academic publishing and the review process. It really helps defuse what tends to be an adversarial relationship with reviewers (or authors).
Reposted by Matthew Muckley
📣 I am sure we have reached only a small fraction of New York's ML community in bsky. Please repost 🔁 this if you think you may have interested people close to you in the social graph.
I did a starter pack of people in New York (City) working on ML/AI. Please distribute and feel free to self nominate!
go.bsky.app/BoEtagz
go.bsky.app/BoEtagz
November 22, 2024 at 2:14 PM
📣 I am sure we have reached only a small fraction of New York's ML community in bsky. Please repost 🔁 this if you think you may have interested people close to you in the social graph.
HELLO Hello hello hellooo...
November 18, 2024 at 4:01 PM
HELLO Hello hello hellooo...