my take on all this llm.c stuff is that it’s very impressive, and karpathy is certainly brilliant, but it was sort of a futile exercise in that llm.c will never be as fast nor as simple as pytorch this whole movement to “write everything in the lowest language possible” is a tad
LLMS
-
Appreciation for Akshay’s Command R+ Share
By
–
Thanks for the share Akshay, Command R+ is really cool!
-
Fine-tuning Open Source LLMs to Outperform GPT-4
By
–
There's a common belief that #opensource LLMs aren't as performant as #GPT4. Well, oftentimes they're not… unless you #finetune! Join our upcoming virtual workshop to learn how you can fine-tune OSS #LLMs to outperform GPT-4: https://
pbase.ai/3xAPXuV -
Snorkel Custom: Enterprise AI Model Deployment Accelerator
By
–
Deliver production AI models within weeks. Introducing Snorkel Custom, an accelerator for enterprises to deliver production AI. Snorkel experts work with you to rapidly evaluate, select, tune, and optimize LLMs for your unique use cases.
Learn more: https://
snkl.ai/custom -
Distilling Step-by-Step: Training Smaller LLMs with Reasoning
By
–
Jason Fries discusses "distilling step-by-step." This method prompts an LLM to provide not just an answer, but also the reasoning behind it. This information is then used to train a smaller, model with similar performance. Watch here: https://
youtube.com/watch?v=TIqf4L
MNCjU
… -
Does scaling solve creative problem solving limitations in AI?
By
–
Do you think that the difficulties with creative problem solving result from insufficient scaling?
-
10 ChatGPT prompts that’ll make you rich
By
–
10 ChatGPT prompts that'll make you rich: (Instant bookmark)
-
Model Parameters and Training Data Reproduction in AI
By
–
Still, this particular example merely demonstrates that a model with more parameters is better at reproducing the training data
-
GPT-4 identifies unreleased game from screenshots without internet
By
–
When given screenshots of an unreleased game that was just revealed this week (that I am very much looking forward to), GPT-4 is able to make some pretty impressive educated guesses, including the game's title, without access to the internet.
-
Deterministic Attention Heads Improve AI Model Capability
By
–
I assure that I would be a lot more capable if I had more deterministic attention heads operating on a larger parameter space