Comments (6)
One idea for this is that you could allow up to two local models to be loaded and assigned to one or more agents. We could load one model into the GPU and the other into the CPU with some RAM allocated to it. So say llama2 into the gpu and use it for most of the agents, and then a Python optimized smaller model into cpu for the engineer agent.
from chatdev.
In theory a long running process could:
- accumulate a queue of prompts/tasks for a specific agent
- swap the engine to that agent and load it in 10-30s
- perform actions and save the queue results of the agent
- then prepare queues for other agents from results
- repeat at 1.
This would allow us to use big models more efficiently, not accumulating a lot of time penalty for VRAM loading times on swaps.
from chatdev.
@andraz @j-loquat your solution and suggestions are looking good to implement.
from chatdev.
Hello, regarding the use of other GPT models or local models, you can refer to the discussion on our GitHub page: #27. Some of these models have corresponding configurations in this Pull Request: #53. You may consider forking the project and giving them a try. While our team currently lacks the time to test every model, it's worth noting that they have received positive feedback and reviews. If you have any other questions, please don't hesitate to ask. We truly appreciate your support and suggestions. We are continuously working to improve more significant features, so please stay tuned.😊
from chatdev.
This has been referenced in #27
from chatdev.
One thing to consider with local LLM agents is that we should keep the prompts shorter than for OpenAI and reduce the temperature to perhaps lower than 0.5. Lower temp and shorter prompts makes a huge difference in local response times as per GPT4All project.
from chatdev.
Related Issues (20)
- _"ghcr.io/wxw-matt/devcontainer-features/script_runner:0": {}_
- Tiltoldum
- KeyError: 'web_spider' when trying to run run.py HOT 2
- Requirements text file missing bs4 HOT 2
- [Feature Request] Attaching files to define task, bootstrap org and name, and for including reference files. HOT 2
- error No module named 'bs4' HOT 1
- 如何配置azure openai接口? HOT 1
- Log Visualizer doesn't show anything in home page HOT 2
- Privacy policy: Does it send, store, or use the prompts and responses anywhere else except the direct LLM endpoints? HOT 1
- 对安装指南的补充 HOT 2
- Rate limiting error HOT 6
- can not create django project which need sub folder structure. HOT 5
- stop should either be excluded, set to [] or utilized. Setting it to null caused errors HOT 1
- Poker is bogus HOT 1
- Report a bug when you switch to use third base url
- The system cannot find the path specified in Install Dependencies: HOT 7
- Issues related with OpenAI_Usage_Info and ChatDev Software Info HOT 4
- Add ollama integration instruction in the readme . HOT 3
- Is it possible to perform incremental development on existing projects that are not generated by ChatDev HOT 3
- cannot further optimize the generated software result. HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from chatdev.