Filip's World
my notes on software, cryptocurrency, and ML


Please use $MODELPATH

The Problem On-device AI models are becoming more common, but by default, applications store them in different places. Application Default How to Change HuggingFace Transformers ~/.cache/huggingface/hub Set TRANSFORMERS_CACHE LM Studio ~/.cache/lm-studio/models My Models -> Change Ollama ~/.ollama Set OLLAMA_MODELS Jan ~/jan/models Advanced Settings -> Jan Data Folder GPT4All ~/.cache/gpt4all/ Unclear These applications (and new ones which will come) are becoming popular with non-technical audiences. A user who tries one of these applications is more likely to try another.

Bridging in Juicebox v4

Juicebox v4 introduces the JBSucker contracts for bridging project tokens and funds (terminal tokens) across EVM chains. Here’s what you’ll need to know if you’re building a frontend or service which interacts with them.

Local LLMs

I wrote this guide for a friend interested in running large language models locally in September 2023, and parts of it are out of date – things change quickly. If you’re reading it and run into questions feel free to email me: f (at)

Back to top