Sandeep @sandeep1337

Data & Pretraining @MistralAI | Previously @NVIDIA | PhD @MILAMontreal | Masters @SCSatCMU | Intern at @MSFTResearch @facebookai @element_ai @DescriptApp sandeeplearning.com San Francisco, CA Joined February 2009

Tweets

65
Followers

2K
Following

596
Likes

1K

Sandeep @sandeep1337

8 months ago

We just dropped new 24B base and instruct models under the Apache 2.0 license. Pretraining this model was a ton of fun!

Mistral AI @MistralAI

8 months ago

We just dropped new 24B base and instruct models under the Apache 2.0 license. Pretraining this model was a ton of fun!

307 492 5K 878K 931

1 0 3 490 0

lmarena.ai @arena

a year ago

The new Mistral Large 2 is here🔥 It's now leading the Arena hard leaderboards. Super impressive performance in Coding, Hard Prompts and Math, surpassing top-tier models GPT-4 Turbo/Claude Opus! Plus, it's an open-weight model -- another great news for the open-source…

Guillaume Lample @ NeurIPS 2024 @GuillaumeLample

a year ago

19 69 510 119K 79

Download Image

9 65 456 81K 102

Download Image

Guillaume Lample @ NeurIPS 2024 @GuillaumeLample

a year ago

Today, we release Mistral Large 2, the new version of our largest model. Mistral Large 2 is a 123B-parameter model with a 128k context window. On many benchmarks (notably in code generation and math), it is superior or on par with Llama 3.1 405B. Like Mistral NeMo, it was trained…

Mistral AI @MistralAI

a year ago

162 366 2K 788K 357

47 269 2K 534K 474

Sandeep @sandeep1337

a year ago

We released Mistral-NeMo 12B today. A model trained with some of my previous collaborators @nvidia. - 128K sequence length - 128k vocab size with good compression in 100+ languages. - Base & FP8 Instruct Model with strong results - Apache 2.0 Blog: mistral.ai/news/mistral-n…

1 3 35 3K 5

Mistral AI @MistralAI

a year ago

mistral.ai/news/mistral-n…

95 189 1K 257K 248

Sandeep @sandeep1337

a year ago

Excited to share one of the things that I've been working on for the past few months after joining Mistral - Codestral-22B. Particularly satisfying to build a code assistant with FIM since it is what I use LLMs for the most. Try it for free on LeChat - chat.mistral.ai/chat

2 3 34 3K 5

Download Image

Sandeep @sandeep1337

a year ago

We just dropped the 8x22B Base and Instruct Models on @huggingface with an Apache 2.0 License! Looking forward to seeing what the open-source community cooks with this! Blog: mistral.ai/news/mixtral-8… HF base: huggingface.co/mistralai/Mixt… HF instruct: huggingface.co/mistralai/Mixt…

1 5 40 3K 2

Mistral AI Labs @MistralAILabs

2 years ago

Thrilled to announce the first @MistralAI Hackathon in San Francisco on March 23-24! Sign up at: partiful.com/e/Zk9c9HVsmtsG… Keynote from Mistral AI founders: @arthurmensch & @GuillaumeLample. Mistral AI mentors: @dchaplot, @sandeep1337, @theo_gervet, @mjmj1oo, @sophiamyang

17 47 316 136K 86

Download Image

Sandeep @sandeep1337

2 years ago

Mistral Large (mistral.ai/news/mistral-l…) is now available at chat.mistral.ai/chat!

0 2 26 5K 3

Arthur Mensch @arthurmensch

2 years ago

We’re announcing a new optimised model today! Mistral Large has top-tier reasoning capacities, is multi-lingual by design, has native function calling capacities and a 32k model. The pre-trained model has 81.2% accuracy on MMLU. Learn more on mistral.ai/news/mistral-l…. Mistral…

84 319 2K 281K 420

Devendra Chaplot @dchaplot

2 years ago

We just released Mixtral 8x7B paper on Arxiv: arxiv.org/abs/2401.04088

45 473 3K 435K 1K

Download Image

Aran Komatsuzaki @arankomatsuzaki

2 years ago

Retrieval meets Long Context Large Language Models Llama with 32K context using retrieval-augmentation at generation outperforms finetuned LLM with 32K context via positional interpolation, while taking much less computation arxiv.org/abs/2310.03025

2 48 209 30K 86

Download Image

Sandeep @sandeep1337

4 years ago

Defended my PhD thesis "On Representation Learning for Generative Models of Text" @Mila_Quebec last week. Shoutout to my awesome committee @sleepinyourhat @chrisjpal, Yoshua Bengio, @AaronCourville, and Philippe Langlais. Excited about my next adventure as a researcher @NVIDIAAI.

8 5 172 0 7

RBC Borealis @RBCBorealis

5 years ago

Thrilled to share the winners of our 2020 Graduate Fellowship Awards. These awards are given to students who've shown outstanding #research capabilities and represent leading Canadian universities & #AI institutes. #Canada Read more: bit.ly/2MjLySh

0 8 30 0 1

Download Image

Shrimai @shrimai_

6 years ago

Our paper titled “Towards Content Transfer through Grounded Text Generation” with @MSFTResearch co-authors @chris_quirk and Michel Galley has been accepted to #naacl2019. Paper: arxiv-export-lb.library.cornell.edu/pdf/1905.05293 Dataset & Code: github.com/shrimai/Toward… Project Page: microsoft.com/en-us/research…

0 20 80 0 8

Sandeep @sandeep1337

6 years ago

If you're at #iclr2019, come check out our poster from 11-1 on Thursday (with @FacebookAI and @milamontreal) on text rewriting arxiv.org/abs/1811.00552. With collaborators @GuillaumeLample, Eric Smith, @LudovicDenoyer, Marc'Aurelio and Y-Lan.

0 3 19 0 0

Bored Yann LeCun @boredyannlecun

7 years ago

And now, starting at small forward..... AAAAAARONNNNNN ...... COOOOOOURVILLLLLLLLE!!!! #torched

Ian Osband @IanOsband

7 years ago

And now, starting at small forward..... AAAAAARONNNNNN ...... COOOOOOURVILLLLLLLLE!!!! #torched

2 4 38 0 1

0 3 23 0 0

Farza 🇵🇰🇺🇸 @FarzaTV

7 years ago

Introducing PogChampNet: a deep convolutional net that takes only a video of Overwatch gameplay as input and can automatically extract the "highlight worthy" moments. Automatic montages! It was trained on data powered by Twitch chat and its memes. blog: goo.gl/KRFYZE