[#Article] @OpenAI proposes free fine-tuning for GPT-4o mini for two months https://actuia.com/actualite/open ai-propose-un-fine-tuning-gratuit-pour-gpt-4o-mini-pendant-deux-mois/ … #AI #ArtificialIntelligence
LLMS
-
HuggingFace MINT-1T Model Collection Released
By
–
here https://
huggingface.co/collections/ml
foundations/mint-1t-6690216ca4d0df7e518dde1c
… -
Meta Releases Llama 3.1 405B Model for Market
By
–
Meta has released Llama 3.1 405B, which is now available on the market. According to the company, with the release of the 405B model, they are poised to supercharge innovation—with unprecedented opportunities for growth and exploration. They believe that the latest generation of
-
Hugging Face and NVIDIA Deploy Open Source AI Models
By
–
Latest open-source models (all sizes, specialized & customized) + optimized hardware & infra = Excited about this release with our friends from @nvidia
! Just tap "Deploy" on models from HF to check your options. https://
venturebeat.com/ai/hugging-fac
e-offers-inference-as-a-service-powered-by-nvidia-nim/
… -
LLMs capture only 1% of human behavior and data
By
–
I would say that your inbox captures max 1% of you. In other words, if you trained LLM on your inbox data, it would predict only 1% of your actions. I think there's so much that is not in your inbox. However, my cryptic tweet was meant to be mainly about setting boundaries on
-
FP16 vs BF16: Precision Mismatch Concerns in Model Fine-tuning
By
–
fp16 or bf16? I’m always a little nervous seeing people finetune or inference in fp16 models that were pretrained in bf16. The number of exponent bits (and hence range) is lower? I have a todo to look into it closer. Depends on the checkpoint possibly.
-
Artificial Intelligence Systems Require Massive Training Data
By
–
You know a system has little intelligence when it takes a trillion words to train it.
-
Reproducing GPT-2 124M Language Model Implementation
By
–
Let's reproduce GPT-2 (124M) https://
bit.ly/3Wi3hOw #AI #MachineLearning #DeepLearning #LLMs #DataScience
