Alexander Dibrov

I was going to gather some results to include with the pull request, but the move to new weights broke my llama-13b-4bit. My llama-7b has attention deficit (and with a...

### **Dual Captioning Test** Images were sent to llama-13b-4bit-128g running with the "LLaMA's primary function is to interact with users..." character card loaded and no chat history. **Relevant parameters:** ```...

Reopening the branch, as I closed it by accident. Was trying to clean things up before adding a code update, but I'm clearly still working my way through the GitHub...

So, renaming branches does not work as I expected - just wanted a more descriptive name than 'patch-1 for it. Thank you for bearing with me.

### **Image Strips and Plain Language Wrapper** I've been able to further improve results using procedures loosely inspired by [retinotopic mapping](https://en.wikipedia.org/wiki/Retinotopy): **1.** Split input into two horizontal strips and two...

### **DeepDanbooru Integration and UI Options** This version of the extension can now use [DeepDanbooru](https://github.com/KichangKim/DeepDanbooru) tagging to help generate prompts, and it can do so in any combination with the...

I would love to see LangChain support. The application I'd be most interested in is [chatbot memory](https://www.pinecone.io/learn/langchain-conversational-memory/). As it stands, I've written a memory extension that does something resembling their...

This is perhaps not as fast as performing all model computation on GPU, but you can get a substantial boost to generation rate if you compile llama.cpp with CUBLAS enabled...

I struggled to do this myself, and steps will vary from system to system, so your mileage may vary. That being said, you can try something like this: **1.** Obtain...

Attached is a rather lengthy example of the sort of output I get with this PR. More than anything, it illustrates the difficulties that my local model has staying on...