1.5 KiB
🐛 #615 - Bug: Gemma3 Vision not working
| Author | erazortt |
|---|---|
| State | ❌ Closed |
| Created | 2025-07-15 |
| Updated | 2025-07-19 |
Description
What happened?
Using the follollowing command the vision is not working: llama-server.exe -m models/gemma-3-27b-it.gguf --mmproj models/gemma-3-27b-it-mmproj-bf16.gguf --temp 1 --top-k 64 --top-p 0.95 --min-p 0.01 -ngl 63 -c 32768 -ctk q8_0 -ctv q8_0 --flash-attn --no-kv-offload --port 10000
When using exactly the same command line on llama.cpp it works.
Name and Version
$ ./llama-server.exe --version
version: 1 (8c2a6ee)
built with MSVC 19.44.35211.0 for
What operating system are you seeing the problem on?
Windows
Relevant log output
💬 Conversation
👤 jmcook commented the 2025-07-16 at 03:36:05:
That's funny, I was just trying the same thing tonight and noticed the same thing!
👤 ikawrakow commented the 2025-07-16 at 09:18:39:
Sorry, there is no vision support in ik_llama.cpp at all. As I know nothing about vision or multi-modality, my suggestion is to try to convince @ngxson to contribute the multi-modality library he created for llama.cpp also to ik_llama.cpp.
👤 ikawrakow commented the 2025-07-19 at 09:27:13:
I think I'll close this one. A feature request can be opened instead.