Apple builds a slimmed-down AI model using Stanford, Google innovations The phone giant's open-source large language model beats previous models by melding the insights of many researchers. https://
zdnet.com/article/apple-
builds-a-slimmed-down-ai-model-using-stanford-google-innovations/
… @Apple
OPEN SOURCE
-
Apple Develops Optimized AI Model Using Stanford Google Research
By
–
-
Perplexity launches Sonar Large 32k on LLaMa 3 70b
By
–
BREAKING: Perplexity released a new Sonar Large 32k model trained on LLaMa 3 70b. Its knowledge cutoff is stated to be April 2024.
-
$200 Bounty: Custom Collator for Unsloth AI Completions
By
–
I'll give $200 to the first person who writes a fully working custom collator for @UnslothAI that allows for only training on completions.
-
Meta’s Llama 3 Advances Open Source LLM Ecosystem
By
–
Incredibly excited to see OSS LLMs leap forward w/ Meta's Llama 3, and for Snorkel AI to be part of the "Llama ecosystem." Thanks to Joe Spisak for the great mention! Watch the full video here: https://
buff.ly/49YQfJv #Llama3 #SnorkelAI -
User seeks feedback on unsloth tool integration blocker
By
–
Hey @danielhanchen just wanted to check on this — it's the biggest blocker for me at the moment. Really excited to try unsloth!
-
LMSYS Production Configuration Transparency Open Source
By
–
Yeah possibly? I don't know much about how lmsys are configured – wish their exact production configuration was open source!
-
Decentralized Social Infrastructure: Open Source Without Vendor Lock-in
By
–
There’s a big difference in switching costs though.
We provide “social plumbing” and infrastructure (and we do hope to be successful as a company!) but the content of the community is just a git clone away, it’s not in some walled garden -
Predibase Launches 10x Faster Fine-tuning Stack with Llama3 Support
By
–
Announcement: New Improved #Finetuning Stack—10x Faster Training! Here are the highlights: New #training stack up to 10x faster + better model quality #Llama3 available for inference + fine-tuning New Python #SDK: More consistent and robust
