aifeifei798

/

Darkidol-Gemma-4-E4B-it

like 12

Hi! Thank you for the support and enthusiasm!

Regarding the Ollama version: I have personally tested it, and currently, Ollama cannot run this model yet.

The reason is that this model uses the brand-new Gemma 4 architecture, and Ollama’s internal engine hasn't been updated to support it yet (it will throw an "unknown model architecture" error).

I've tried creating an Ollama Modelfile, but it currently fails with a 500 Internal Server Error because Ollama’s backend is not yet synced with the latest llama.cpp code for gemma4.

To run this model right now (including Vision features), please use the latest llama-server from llama.cpp with the following command:

./llama-server -m Darkidol-Gemma-4-E4B-it.Q8_0.gguf --mmproj Darkidol-Gemma-4-E4B-it.mmproj-f16.gguf

You can download the necessary GGUF files here:

• Main Model: https://huggingface.co/mradermacher/Darkidol-Gemma-4-E4B-it-GGUF
• i1-GGUF Version: https://huggingface.co/mradermacher/Darkidol-Gemma-4-E4B-it-i1-GGUF

I will release the Ollama manifest and push it to the library as soon as Ollama officially adds support for Gemma 4. Thanks for your patience!