Ollama

Explaining the concept of K/V context cache quantisation, why it matters and the journey to integrate it into Ollama. Why K/V Context Cache Quantisation Matters The introduction of K/V context cache quantisation in Ollama is significant, offering users a range of benefits: • Run Larger Models: With reduced VRAM demands, users can now run larger, more powerful models on their existing hardware. • Expand Context Sizes: Larger context sizes allow LLMs to consider more information, leading to potentially more comprehensive and nuanced responses. For tasks like coding, where longer context windows are beneficial, K/V quantisation can be a game-changer. • Reduce Hardware Utilisation: Freeing up memory or allowing users to run LLMs closer to the limits of their hardware. Running the K/V context cache at Q8_0 quantisation effectively halves the VRAM required for the context compared to the default F16 with minimal quality impact on the generated outputs, while Q4_0 cuts it down to just one third (at the cost of some noticeable quality reduction). ...

Gollama on Github Gollama is a client for Ollama for managing models. It provides a TUI for listing, filtering, sorting, selecting, inspecting (coming soon!) and deleting models and can link Ollama models to LM-Studio. The project started off as a rewrite of my llamalink project, but I decided to expand it to include more features and make it more user-friendly. ...

Confuddlement on Github I was tired of manually downloading Confluence pages and converting them to Markdown, so I wrote a small command-line tool designed to simplify this process. Confuddlement is a Go-based tool that uses the Confluence REST API to fetch page content and convert it to Markdown files. It can fetch pages from multiple spaces, skip pages that have already been fetched, and summarise the content of fetched pages using the Ollama API. $ go run ./main.go Confuddlement 0.3.0 Spaces: [COOLTEAM, MANAGEMENT] Fetching content from space COOLTEAM COOLTEAM (Totally Cool Team Homepage) Retrospectives Decision log Development Onboarding Saved page COOLTEAM - Feature List to ./confluence_dump/COOLTEAM - Feature List.md Skipping page 7. Support, less than 300 characters MANAGEMENT (Department of Overhead and Bureaucracy) Painful Change Management Illogical Diagrams Saved page ./confluence_dump/Painful Change Management.md Saved page Illogical Diagrams to ./confluence_dump/Ilogical Diagrams.md Done! $ go run ./main.go summarise Select a file to summarise: 0: + COOLTEAM - Feature List 1: + Painful Change Management 2: + Illogical Diagrams Enter the number of the file to summarise: 1 Summarising Painful Change Management... "Change management in the enterprise is painful and slow. It involves many forms and approvals." go run main.go -q 'who is the CEO?' -s 'management' -r 2 Querying the LLM with the prompt 'who is the CEO?'... "The CEO of the company is Peewee Herman." Usage Running the Program Copy .env.template to .env and update the environment variables. Run the program using the command go run main.go or build the program using the command go build and run the resulting executable. The program will fetch Confluence pages and save them as Markdown files in the specified directory. Querying the documents with AI You can summarise the content of a fetched page using the Ollama API by running the program with the summarise argument: go run main.go summarise To perform a custom query, you can use the query argument: -q: The query to to provide to the LLM. -s: The search term to match documents against. -r: The number of lines before and after the search term to include in the context to the LLM. go run main.go -q 'who is the CEO?' -s 'management' -r 2 Querying the LLM with the prompt 'who is the CEO?'... "The CEO of the company is Peewee Herman." ...

“Should I run a larger parameter model, or a higher quality smaller model of the same family?” TLDR; Larger parameter model [lower quantisation quality] > Smaller parameter model [higher quantisation quality] E.g: Qwen2.5 32B Q3_K_M > Qwen2.5 14B Q8_0 Caveats: Don’t go lower than Q3_K_M, or IQ2_M, especially if the model is under 30B~ parameters. This is in the context of two models of the same family and version (e.g. Qwen2.5 Coder). Longer answer: Check out the Code Chaos and Copilots slide deck. ...