Zhao fuyong

Problem in first phase of training:Token {'<Img>'} is not a string but a <class 'set'>

2

When training with part of the cc3m dataset, the following error was encountered for cc3m-only training: TypeError: Token {''} is not a string but a . 

The cc3m raw dataset 1/3 link is dead, has any of you downloaded the full version of the dataset? Is it possible to get a backup copy of it?

I may not need to train all the functions in the actual requirement, is it possible to keep only the text+image output function on the output side?

Finetuned MiniCPM-V2.5 erro

4

I'm trying to fine-tune MiniCPM-V2.5, but I get the following error: f"{','.join(outputs.keys())}. For reference, the inputs it received are { ^ SyntaxError: unterminated string literal (detected at line 38)

model_type: 'minicpm-v-v2_5-chat' is not registered.

7

I do pip install 'ms-swift[llm]' -U and still get an error: model_type: 'minicpm-v-v2_5-chat' is not registered.