Teknium

ShareGPT Dataset?

8

Hello, I see ShareGPT's dataset is listed on the readme, but the download for alpaca format version is not listed. Can it get listed? Very interested.

Any chance we could improve the dataset beyond fixing?

41

Would that be relevant in the scope of this project? Like adding a couple sorts of task examples could improve its generalized capabilities, for instance: Longer responses GPT-4 Generated Responses...

Slower tokens/s than expecting

14

Hello I am running a 2x 4090 PC, Windows, with exllama on 7b llama-2. I am only getting ~70-75t/s during inference (using just 1x 4090), but based on the charts,...

## Why are these changes needed? In model_adapters, llama-3 supercedes Hermes-2-Pro-Llama-3 and Hermes-2-Theta-Llama-3 when detecting it's format, so I changed it to have `meta-llama` as its match string to make...

How do you get the model to be good at code if it downsamples code?

1

Question in topic..