Aran Komatsuzaki@arankomatsuzaki

2 years ago

Nvidia just released the weights of NV-Embed-v1, the current leader on MTEB. Some additional info below: Base model: Mistral-7B-v0.1 Pooling type: Latent-Attention Embedding dimension: 4096 Max input tokens: 32k https://t.co/mTH5wwxtjO

NV-Embed: NVIDIA’s Groundbreaking Embedding Model Dominates MTEB Benchmarks #DL #AI #ML #DeepLearning #ArtificialIntelligence #MachineLearning #ComputerVision #AutonomousVehicles #NeuroMorphic #Robotics https://t.co/E5SYh5KgJD

🙋🏻‍♂️@NVIDIAAI just released a cool EMBEDDING model , that's "instruction" based ! that means you can use it with @llama_index instructor embeddings function, or try it out on @huggingface directly with my little @Gradio demo : https://t.co/2FU3MwLwU2 this is the future !