ryan-yang125 / chatllm-web Goto Github PK

要怎么切换成自定义模型，比如ChatGLM-6B？

The line break feature can be triggered by shift + enter? This will give us a more fluent prompt writing experience

Had to change directories to run the npm i - patch the readme to include the cd.

模型文件能不能下载到本地从本地加载，从huggingface.co下载有点慢，而且过一段时间浏览器清除缓存之后又要重新下载。
能否给个资源包放在开发服务器里，然后修改配置从本地加载

Logs:
Initialize GPU device: WebGPU - NVIDIA GeForce RTX 3060
Initialize GPU device: WebGPU - NVIDIA GeForce RTX 3060
Initialize GPU device: WebGPU - NVIDIA GeForce RTX 3060
Initialize GPU device: WebGPU - NVIDIA GeForce RTX 3060
[System Initalize] Loading GPU shader modules[54/54]: 100% completed, 10 secs elapsed.
[System Initalize] All initialization finished.
[System Initalize] All initialization finished.
[System Initalize] All initialization finished.
[System Initalize] All initialization finished.
[System Initalize] All initialization finished.
Generate error, OperationError: The operation failed for an operation-specific reason
Generate error, OperationError: The operation failed for an operation-specific reason
Generate error, OperationError: The operation failed for an operation-specific reason
Generate error, OperationError: The operation failed for an operation-specific reason
Generate error, OperationError: The operation failed for an operation-specific reason

use @mlc-ai/web-llm to refactor code

Thanks for building this! Desktop PWA support would be useful. Looking forward to using this more.

support local document search https://github.com/matthoffner/web-llm-embed
support multi models such as llama2
save prompt
agent support https://github.com/idosal/AgentLLM

how can I choose my Nvidia GPU rather than intel graphics gpu, its really slow when I use default gpu, and I choose Nvidia in Nvidia control panel for google chrome canary, it still uses intel gpu.

setting page about:

1. model:

• models selection
• param config: temperature, max-length

2. device:

• gpu switch
• cache usage

3. ui

• dark/light theme

I tried to add the DoctorGPT model, modified the model in /public/lib/vicuna-7b, and also modified the config.json to point the cacheUrl to the local model at http://localhost:3000/lib/WebLLM/vicuna-7b/doctorGPT/.

However, the webpage shows an error：