would you draw a wider conclusion such as "hardware portability"?
I read the paper and concluded "JAX + TPU is unbeatable". It's hard to conclude anything more general.
The paper doesn't cover any other set of hardware — AMD, Trainium, GraphCore, etc.
AI
-
JAX TPU Performance Limits Hardware Portability Claims
By
–
-
Libraries and tech stack recommendations
By
–
Would love to know the libraries/tech stack you’re working with. 🙂
-
Web Automation and Scraping Experts on X
By
–
So who are the amazing folks on X working on Web Automation and Web Scraping projects? Tag them below or say a
-

Baichuan 3: 100B+ Parameter LLM Surpasses GPT-4 Chinese
By
–
Baichuan AI released LLM Baichuan 3 with parameters exceeding 100 billion. It has substantial improvements in foundational capabilities, surpassing GPT-4 in various authoritative evaluations of Chinese tasks and emerging as the top-performing LLM for Chinese medical tasks.
-
AI Models Scale Exponentially: Trillions Parameters Nine Orders Magnitude
By
–
In just a few years, cutting-edge models have gone from using millions of parameters to trillions. The amount of computation used to train the largest AI models has increased by nine orders of magnitude. Here's what we should do next:
-
General Purpose Technologies: Magic Propelling Civilization Forward
By
–
General purpose technologies are the closest thing we have to magic. Each has given us more with less – and propelled civilization forward. Check out this list of 24 over the last 10,000 years – Domestication of plants – Domestication of animals – Smelting of ore – Money –
-

Building Chatbots with Source Citations Using LangChain
By
–
Question Answering with Citations When building a chatbot over documents, it's often useful to ask it to cite it's sources. This can help build trust and check responses We've recently added a detailed guide on how to do this! Guide: https://
python.langchain.com/docs/use_cases
/question_answering/citations
… -

AGI Orchestrator Pass: Web3 AI Utility for NFT Collectors
By
–
[ Register Your AGI ORCHESTRATOR PASS ] The AGI ORCHESTRATOR is an experimental Web3 AI/AGI utility learning to orchestrate synergies amongst #NFT Collectors (under development) Mint for 250 $AGI : https://
montrealai.xyz/o.html Pre-Alpha Version #AGIFirst #AGIOrchestrator #ENS -

Codellama-70b-instruct success rate, prompt variations
By
–
FWIW I wasn’t able to reproduce, using http://
labs.perplexity.ai with codellama-70b-instruct. I guess it depends on some detail of system prompt / format. I got 10/10 success using both prompts from the replies here and my own. -

MLX Community Releases Quantized Code Llama Models
By
–
Just finished uploading 4 MLX models. Quantized Code Llama 7B and 13B, Python and Instruct! Link here https://
huggingface.co/mlx-community