Why is the output of the following model so different?
" In this image, there is ", which is a prompt. When I input this prompt into the model, the model can output normally. However, if the prompt I give to the model is " in this image, what did you see?", the model starts to talk nonsense. Why?
This is a base model, not a chat/instruction tuned model, that might be the reason the completions seem unexpected.
For the instruction tuned model you may want to check https://huggingface.co/google/gemma-3n-E4B-it
Hi @rrrfy ,
Welcome to Google Gemma family of open source models, I can successfully able to get the output for the given two prompts which are mentioned in your comment. I can see similar kind of output is produced by the model for the given prompts. However both outputs are not exactly same, at the same time the content of the both the outputs are similar in nature. It's highly recommend to check whether the downloaded weights got corrupted or any files that are dowloaded corrupted please. Please find the attached gist file for your reference.
Thanks.