* Update README.md * fix: 'ollama run' should be 'ollama pull' * fix: linting, syntax, spelling corrections for all docs * fix: markdown linting rules and missed fixes * fix: readded space to block * fix: changed sh blocks to text * docs: added exception for bare urls in markdown * docs: added exception for in-line html (MD033/no-inline-html) * docs: made python indentation level consistent (4 space tabs) even though I prefer 2. --------- Co-authored-by: Charles Packer <packercharles@gmail.com>
33 lines
1.3 KiB
Markdown
33 lines
1.3 KiB
Markdown
---
|
|
title: koboldcpp
|
|
excerpt: Setting up MemGPT with koboldcpp
|
|
category: 6580da9a40bb410016b8b0c3
|
|
---
|
|
|
|
1. Download + install [koboldcpp](https://github.com/LostRuins/koboldcpp/) and the model you want to test with
|
|
2. In your terminal, run `./koboldcpp.py <MODEL> -contextsize <CONTEXT_LENGTH>`
|
|
|
|
For example, if we downloaded the model `dolphin-2.2.1-mistral-7b.Q6_K.gguf` and put it inside `~/models/TheBloke/`, we would run:
|
|
|
|
```sh
|
|
# using `-contextsize 8192` because Dolphin Mistral 7B has a context length of 8000 (and koboldcpp wants specific intervals, 8192 is the closest)
|
|
# the default port is 5001
|
|
./koboldcpp.py ~/models/TheBloke/dolphin-2.2.1-mistral-7B-GGUF/dolphin-2.2.1-mistral-7b.Q6_K.gguf --contextsize 8192
|
|
```
|
|
|
|
In your terminal where you're running MemGPT, run `memgpt configure` to set the default backend for MemGPT to point at koboldcpp:
|
|
|
|
```text
|
|
# if you are running koboldcpp locally, the default IP address + port will be http://localhost:5001
|
|
? Select LLM inference provider: local
|
|
? Select LLM backend (select 'openai' if you have an OpenAI compatible proxy): koboldcpp
|
|
? Enter default endpoint: http://localhost:5001
|
|
...
|
|
```
|
|
|
|
If you have an existing agent that you want to move to the koboldcpp backend, add extra flags to `memgpt run`:
|
|
|
|
```sh
|
|
memgpt run --agent your_agent --model-endpoint-type koboldcpp --model-endpoint http://localhost:5001
|
|
```
|