will not support MiniCPM-V-4 in the future version?

This would be so cool! @jmorganca could someone review this PR and merge?

…hen-there-is-a-slice-image

fix bug

I’m bringing MiniCPM-V 4.5 in this PR.

Compared to the recently released MiniCPM-V 4.0, this new version puts more emphasis on performance metrics.

good model!!
Is there any possibility to add tools?

Thank you

We'll pick this up next time we do a vendor sync but we are unable to do one off updates for models like this.

I’m bringing MiniCPM-V 4.5 in this PR.
Compared to the recently released MiniCPM-V 4.0, this new version puts more emphasis on performance metrics.

good model!! Is there any possibility to add tools?

Thank you

What does "add tools" mean? Could you please explain it to me in more detail? Perhaps I can help you out.

We'll pick this up next time we do a vendor sync but we are unable to do one off updates for models like this.

I'd like to ask, how often does the Ollama team "vendor sync"? Or is it a fixed time every month?

I’m bringing MiniCPM-V 4.5 in this PR.
Compared to the recently released MiniCPM-V 4.0, this new version puts more emphasis on performance metrics.

good model!! Is there any possibility to add tools?
Thank you

What does "add tools" mean? Could you please explain it to me in more detail? Perhaps I can help you out.

The ability to send instructions to exectute a tool, the template and the model must support tool calls

like this
https://github.com/ollama/ollama-python/blob/main/examples/tools.py

We'll pick this up next time we do a vendor sync but we are unable to do one off updates for models like this.

I'd like to ask, how often does the Ollama team "vendor sync"? Or is it a fixed time every month?

It's not fixed but there's one in progress now:
#12245

@jessegross We can discuss this: When we release the MINiCPM-V series models, I'll submit a pull request with minimal changes to ensure the new models run on Ollama. Once you've synchronized the llama.cpp file, I can simply overwrite it. Is that okay?

The MINiCPM-V series is also quite popular in the community. Among the multimodal models in the Ollama model library, it has over three million downloads, ranking it as the third most popular. We hope to quickly adapt the models to Ollama after release so that the community can quickly use them.

The best way to get first class support is to implement minicpm-v directly in the Ollama engine, as I believe we have discussed before. In that case, we would be able to incorporate changes immediately (effectively, you would become the maintainer for the model and we would look to you to ensure the best experience). In addition, it would be able to take advantage of improvements to the engine, such as the new memory management coming in 0.11.11.

We really can't make one off changes for models in the old engine. We would just bring all models updated there as a group and the frequency is not guaranteed as it can take significant time to incorporate and stabilize.

UP. Still checking every day for updates.