yahma

Has anyone tested this yet? We have a use case for our European partners from German speaking countries. Would like to know what other people's experiences are.

Very exciting for multi-lingual models. I really hope this one performs as well as the benchmarks suggest.

Yes. This is known as Mixture of Experts (MOE).

We already have several promising ways of doing this:

1. QMoE: A Scalable Algorithm for Sub-1-Bit Compression of Trillion-Parameter Mixture-of-Experts Architectures. Paper - Github
2. S-Lora: Serving thousands of concurrent adapters.
3. Lorax: Serve hundreds of concurrent adapters.
4. LMoE: Simple method of dynamically loading Loras

But are the short responses more correct?

Game changer! Would love to see this incorporated into ExLLama, AutoGPTQ and LlamaCPP

The startup time makes Replicate nearly unusable for me. Only popular models stay in memory. Other less used models shutdown, and you need to wait for startup before first inference.

Do you have Link to the stl for the 3d print?

Today I learned there are a large group of "EA"s (Effective Altruists) who are composed of millionaires and people in high positions of power.

These people believe it is their duty to act as the 'gatekeepers' of AI and prevent regular people from have useful or powerful AI. They want to destroy the open source AI movement and any company that is willing to allow regular people access to powerful AI.

Would love to use this for handling remote security camera footage.

Tried with LLAVA with little success. Has anyone successfully applied any of the Open Vision models to the problem of security?

None of the open models perform function calling as well as openai...

Dataset release??

Do we get the dataset this time?