Great read! My experience is that you’re fighting physics but also the nvidia compiler and the stack overall, and even after pulling *a lot* of tricks we still can’t achieve more than ~80-90% mem bw on many kernels that you’d naively think should be ~100. And the rabbit hole
CODE
-
Code Predictability Versus Prose Writing Principles
By
–
In writing code the most obvious and predictable thing is usually what you want When writing actual prose it's more or less the opposite!
-
Fine-tuning AI Models on Custom Work Data
By
–
I'd love to see you finetune a model on your own work though I can help
-

7 Steps to Mastering Data Engineering: Essential Roadmap
By
–
7 Steps to Mastering Data Engineering: The only data engineering roadmap you need for an introduction to concepts, tools, and techniques to collect, store, transform, analyze, and model data. https://
kdnuggets.com/7-steps-to-mas
tering-data-engineering?utm_source=dlvr.it&utm_medium=twitter&utm_campaign=7-steps-to-mastering-data-engineering
… -

Fine-tune Llama 3 for Customer Service with Free Credits
By
–
Llama-3 is the new hotness, but if you want to get the most out of the #model, then you need to #finetune. Check our #tutorial to learn how to easily fine-tune #Llama3 for customer service use cases. Notebook + data + $25 in free credits included! https://
pbase.ai/4dzyDGY -

Token counts now available in streaming API
By
–
I’m sure you saw this, Simon, but we shipped token counts in the streaming API this week! Thanks again for this suggestion!
-

LeRobot Open Source Library Tops GitHub Trending Rankings
By
–
…yes ok i know the latest iPad ad is horrible, fully agree… but in other news the OSS lib « lerobot » is at the top of GitHub trending all lang, exciting!
-
LangSmith Enhances Personal Prompt Management and Version Tracking
By
–
🛠️ Personal prompts in LangSmith are now easier to access and manage.
— LangChain (@LangChain) 8 mai 2024
You'll see a new home for your organization's prompts, separate from the LangChain Hub. Organize and pull existing prompts by their commit versions — so you can track the performance of different versions of… pic.twitter.com/WNUtLy5NDlPersonal prompts in LangSmith are now easier to access and manage. You'll see a new home for your organization's prompts, separate from the LangChain Hub. Organize and pull existing prompts by their commit versions — so you can track the performance of different versions of