使用Chat-GLM3开源模型报错

[QingChengLineOne]

trafficstars

我将模型替换为Chat-GLM3，在运行的时候报错：

[better629]

@QingChengLineOne what's the error log and your configuration?

[QingChengLineOne]

@QingChengLineOne what's the error log and your configuration? 这是我的key.yaml配置文件

https://github.com/geekan/MetaGPT/issues/526#issuecomment-1840277708 这是chatclm3的最后输出 这是报错信息

[QingChengLineOne]

@QingChengLineOne what's the error log and your configuration? 这是我的key.yaml的配置文件

[better629]

@QingChengLineOne It seems that the LLM output format which does't follow the instruction format(let llm generate in json format). So it failed when load the output string. If you want to try with open llm models, maybe you can have a try with https://github.com/better629/MetaGPT/tree/feat_repair_llmoutput and https://github.com/geekan/MetaGPT-docs/blob/main/src/zh/guide/tutorials/integration_with_open_llm.md. And the PR is still under CR and not released.

[geekan]

https://github.com/geekan/MetaGPT/pull/500 fix this. Would you plz try again?