Sharad Vikram (@sharadvikram) / X

Sharad Vikram

331 posts

Sharad Vikram

@sharadvikram

Researcher @ Google Deepmind. I work on JAX + Pallas (github.com/jax-ml/jax) and Gemini. In the past I worked on Oryx and TFP. I like learning.

San Francisco

Joined August 2012

Pinned
Sharad Vikram
@sharadvikram
Dec 6, 2023
Built with JAX!
Sundar Pichai
@sundarpichai
Dec 6, 2023
Introducing Gemini 1.0, our most capable and general AI model yet. Built natively to be multimodal, it’s the first step in our Gemini-era of models. Gemini is optimized in three sizes - Ultra, Pro, and Nano Gemini Ultra’s performance exceeds current state-of-the-art results on
88K
Sharad Vikram
@sharadvikram
Oct 17, 2018
Excited to share the LORACs prior, coming out my internship at Google: arxiv.org/abs/1810.06891. We use nonparametric hierarchical clustering priors for VAEs for joint structure and representation learning. w/ Matthew Hoffman and @SingularMattrix
Sharad Vikram
@sharadvikram
Jul 22, 2020
Wrote a blog post about building a rendering engine in JAX! sharadvikram.com/blog/ray_march… Hope you all think it's interesting and let me know what you think!
Sharad Vikram
@sharadvikram
Mar 28, 2024
Grouped matrix multiplication on TPU!
Trevor Gale
@Tgale96
Mar 28, 2024
Replying to @Tgale96
I’m not done with MegaBlocks 😁 @apaszke @epiqueras1 @sharadvikram and I just dropped something we’ve been working on for a bit yesterday. MegaBlocks + JAX + TPU = MegaBlox 🔥 github.com/google/jax/pul…
4K
Sharad Vikram
@sharadvikram
Jan 21, 2024
I can't overstate how much I've learned from Sholto and Enrique (and James ofc) over the last year.
Sholto Douglas
@_sholtodouglas
Jan 21, 2024
Enrique would belong on that list just as much (or more) than I. One of our colleagues once likened him to the engineering equivalent of “that student your grad school physics professor talks about as the best they’ve seen in a decade”.
41K
Sharad Vikram
@sharadvikram
Feb 15, 2024
1M context (and beyond) unlocked!!!
Jeff Dean
@JeffDean
Feb 15, 2024
Gemini 1.5 Pro - A highly capable multimodal model with a 10M token context length Today we are releasing the first demonstrations of the capabilities of the Gemini 1.5 series, with the Gemini 1.5 Pro model. One of the key differentiators of this model is its incredibly long
1.6K
Sharad Vikram
@sharadvikram
Dec 23, 2018
I'm happy to announce that the LORACs prior was accepted into #AISTATS2019!
Sharad Vikram
@sharadvikram
Oct 17, 2018
Excited to share the LORACs prior, coming out my internship at Google: arxiv.org/abs/1810.06891. We use nonparametric hierarchical clustering priors for VAEs for joint structure and representation learning. w/ Matthew Hoffman and @SingularMattrix
Sharad Vikram
@sharadvikram
Mar 1, 2024
Replying to @srush_nlp and @typedfemale
The hope is that ML researchers can write kernels as easily as they can write JAX!
1.4K
Sharad Vikram
@sharadvikram
Mar 28, 2024
Replying to @finbarrtimbers
Can I cite this in my perf?
748
Sharad Vikram
@sharadvikram
Apr 18, 2019
If you're at #AISTATS, come check out my poster (No. 94) about the LORACs prior at the poster session! w/ @SingularMattrix and Matt Hoffman
Sharad Vikram
@sharadvikram
Feb 27, 2019
Cool ICLR paper by Justin Fu et al, coming out of a Google AI internship!
Sergey Levine
@svlevine
Feb 27, 2019
Should agents decode language instructions into actions or rewards? We find that recovering a reward function from an instruction and then optimizing it with RL results in substantially better instruction following arxiv.org/abs/1902.07742 w/ Justin Fu, @sguada, A. Korattikara
Sharad Vikram
@sharadvikram
Mar 8, 2024
Replying to @epiqueras1 and @jekbradbury
Why are there two photos of James there?
572
Sharad Vikram
@sharadvikram
Mar 6, 2024
Replying to @typedfemale
@SingularMattrix should we give it a stab?
562
Sharad Vikram
@sharadvikram
Jan 17, 2019
JAX is awesome! I plan on making it my go-to deep learning library very soon.