Hugging Face just dropped Chat Assistants, the open-source version of OpenAI GPTs. Here's what stands out: 1. It's completely free to use & share (compared to needing a $20 subscription) 2. You can select from any open-source LLMs to run your Assistant on. Choose from Mixtral,
LLMS
-
Evaluating LLM Applications: LangSmith Features
By
–
"Is my LLM application getting better or worse over time?" This is one of the biggest questions that AI engineers struggle with. Evaluation of LLM application is still quite difficult. We've recently released two features in LangSmith to make this easier First, to help
-
MLX LoRA Fine-tuning Guide for Language Models
By
–
https://
github.com/ml-explore/mlx
-examples/blob/main/llms/mlx_lm/LORA.md
… -
Request for MoonDream1 Model Support Implementation
By
–
I love this so much! Can we get support for MoonDream1 by @vikhyatk as well?
-
OLMo: Accelerating the Science of Language Models
By
–
OLMo: Accelerating the Science of Language Models Groeneveld et al.: https://
arxiv.org/abs/2402.00838 #ArtificialIntelligence #DeepLearning #MachineLearning -
LoRAX v0.7 Release: Runtime LoRA Merging Per Request
By
–
Announcing the Release of #LoRAX v0.7: Merge #LoRAs at runtime per request! Use cases: Mix LoRAs to improve response (Mixture of LoRAs) No need to pick the right #LoRA by hand Combine with a ranker and merge top-k LoRAs https://
pbase.ai/42tbst0 -
LLM Hallucinations in Legal Contexts: Study Reveals Significant Risks
By
–
Hallucinations are a big problem with large language models. How pervasive is it in legal contexts? A study involving three popular LLMs raises red flags.
-
Chain-of-Thought Information Density and Compute Scaling
By
–
A key insight from chain-of-thought is around the idea of information density. Language models can only do so much with a single forward pass, and so the amount of compute the language model can use must be scaled proportional to how hard a prompt is to solve. What is
-
Fine-tuning Phi-2 with OpenHermes-2.5 Dataset
By
–
My goal next is to fine-tune phi-2 using the incredible OpenHermes-2.5 dataset from @teknium
. Stay tuned! -
Phi-2-DPO-7K: Microsoft’s Fast Fine-Tuned Language Model
By
–
Introducing phi-2-dpo-7k! A Microsoft Phi-2, fine-tuned on a diverse cocktail of 7k chat interactions from @argilla_io
's latest DPO dataset, including orca pairs, ultra-feedback ratings, and capybara-dpo Runs ultra-fast on Apple Silicon, thanks to MLX