MACHINE LEARNING

→ View original post on X — @miketamir

Understanding Deep Learning Book by Simon Prince Released

By

@miketamir

–

11 December 2023 17h37

GitHub – udlbook/udlbook: Understanding Deep Learning – Simon J.D. Prince https://
bit.ly/3Raje5M #AI #MachineLearning #DeepLearning #LLMs #DataScience

→ View original post on X — @soumithchintala

Batch Size Constraints: GPU SM Saturation and Memory Limits

By

@soumithchintala

–

11 December 2023 17h25

For a large enough batch size on a given expert, you'll either saturate the SMs or run out of memory

→ View original post on X — @jxmnop

BPE Tokenization Dies: 2024 Marks Major Transformer Evolution

By

@jxmnop

–

11 December 2023 17h07

i can see it now: 2024 will be remembered as the year BPE died tokenization is by far the clunkiest part of a transformer; one last remaining bit of inelegance in an otherwise hyperoptimized model architecture time for it to go

→ View original post on X — @googleai

Google Showcases Cutting Edge AI Projects at NeurIPS 2023 Expo

By

@googleai

–

11 December 2023 16h13

The #NeurIPS2023 Expo floor is now open! If you're attending @NeurIPSConf
, stop by the Google booth (#215) where you can chat with researchers, attend a demo or Q&A session, and learn about some of our many cutting edge projects. http://
goo.gle/47OE0Pv

→ View original post on X — @godofprompt

ChatGPT Screenshot to Code GPT converts website screenshots to code

By

@godofprompt

–

11 December 2023 15h44

ChatGPT can be a powerful coding assistant..

That's why I've built Screenshot to Code GPT.

Upload any screenshot of a website, and it will turn it to code in seconds!

Try it here: https://t.co/0dcCMJGR41 #CustomGPTs #GPTs #GPTPlus #GPTStore #ChatGPT4 #ChatGPTPlus #OpenAI… pic.twitter.com/ayqbSvpUV9
— God of Prompt (@godofprompt) 11 décembre 2023

ChatGPT can be a powerful coding assistant.. That's why I've built Screenshot to Code GPT. Upload any screenshot of a website, and it will turn it to code in seconds! Try it here: https://
godofprompt.ai/gpts/screensho
t-to-code-gpt
… #CustomGPTs #GPTs #GPTPlus #GPTStore #ChatGPT4 #ChatGPTPlus #OpenAI

Mixtral Outperforms Llama 2 70B on European Language Benchmarks

By

–

11 December 2023 15h20

Mixtral has been trained on a lot of multilingual data and significantly outperforms Llama 2 70B on French, German, Spanish, and Italian benchmarks. (4/n)

Mixtral outperforms Mistral 7B in science, mathematics, and code

By

–

11 December 2023 15h20

Compared to Mistral 7B, Mixtral is significantly stronger in science, in particular in mathematics and code generation. (5/n)

Mixtral Architecture: 8 Feedforward Blocks with Router-Selected Experts

By

–

11 December 2023 15h20

Mixtral has a similar architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks. For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. (2/n)

Mixtral: 12B Speed with 45B Parameter Access via Expert Selection

By

–

11 December 2023 15h20

Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, Mixtral decodes at the speed of a 12B model, while effectively having access to 45B parameters. (3/n)