[Question]: Why ragflow need to use a model named "qwen2-instruct-1-0" when the Chunk Method is "Knowledge Graph"？

[wuxiweili]

Describe your problem

My default model is named "qwen2-instruct" in System. Why ragflow need to use a model named "qwen2-instruct-1-0" when the Chunk Method is "Knowledge Graph"？

Error: xinference.api.restful_api 2829 ERROR [address=0.0.0.0:43828, pid=2863] Model not found, uid: qwen2-instruct-1-0

[SummerDream233]

I find you have implement RAGFlow in your computer. Therefore, you need to write GPT API in your system setting. In the version of online demo, each user will obtain some free tokens from GPT. When it comes to local version, you must prepare API key to run the RAGFlow

[KevinHuSh]

The error message is from xinference. Maybe the model in xinference is named 'qwen2-instruct-1-0'. RAGFlow will not rename the model.

[Muttermal]

我也遇到了同样的问题。我用xinference部署的bge-m3和glm4-chat

但在ragflow中模型名称都变了，加了"-1-0"的后缀:

[skytodmoon]

mark.met the same problem, when I use the graphrag mode:`ERROR: Error code: 400 - {'detail': '[address=0.0.0.0:50218, pid=40327] Model not found, uid: qwen2.5-instruct-2-0'}

{'root': {}} ERROR: Error code: 400 - {'detail': '[address=0.0.0.0:50218, pid=40327] Model not found in the model list, uid: qwen2.5-instruct-2-0'}

{'root': {}} ERROR: Error code: 400 - {'detail': '[address=0.0.0.0:50218, pid=40327] Model not found, uid: qwen2.5-instruct-2-0'} ` but actually in xinference and database, it is qwen2.5, where is the address varible stored?

[skytodmoon]

mark.met the same problem, when I use the graphrag mode:`ERROR: Error code: 400 - {'detail': '[address=0.0.0.0:50218, pid=40327] Model not found, uid: qwen2.5-instruct-2-0'}

{'root': {}}

ERROR: Error code: 400 - {'detail': '[address=0.0.0.0:50218, pid=40327] Model not found in the model list, uid: qwen2.5-instruct-2-0'} {'root': {}} ERROR: Error code: 400 - {'detail': '[address=0.0.0.0:50218, pid=40327] Model not found, uid: qwen2.5-instruct-2-0'} ` but actually in xinference and database, it is qwen2.5, where is the address varible stored?

I restart the xinference llm service and the problem fixed.

[Mar2812]

@infiniflow