mradermacher

/

model_requests

like 140

Description:
Hello @mradermacher ! First, thank you for providing the excellent GGUF quantizations of the Qwen3-VL-32B-Heretic-v2 model. The community really appreciates your work! 😊

Issue:
The current GGUF repository only contains the main model files but is missing the MMProj (multimodal projector) file, which is essential for vision-language functionality.

Current Status:

• ✅ Model files: Available (Q2_K to Q8_0)
• ❌ MMProj file: Missing
• ❌ Vision capability: Not working without MMProj

Request:
Could you please add the MMProj file to enable full vision-language capabilities? The model is based on rpDungeon/Qwen3-VL-32B-Heretic-v2, which should have the necessary components.

Expected File Name:
Qwen3-VL-32B-Heretic-v2.mmproj-f16.gguf (following your naming convention)

Impact:

• Without MMProj: Model works as text-only only
• With MMProj: Full vision-language functionality with zero censorship
• Community Benefit: Complete Heretic 32B experience for local AI users

Alternative Solutions:

1. Extract MMProj from the original HF model
2. Use compatible MMProj from standard Qwen3-VL-32B (if compatible)
3. Provide instructions for users to extract it themselves

Technical Context:

• Base model: rpDungeon/Qwen3-VL-32B-Heretic-v2
• Your GGUF: mradermacher/Qwen3-VL-32B-Heretic-v2-GGUF
• MMProj needed for: Image processing, vision understanding

Community Use Case:
Many users want to use the Heretic 32B model for:

• Uncensored image analysis
• Advanced vision-language tasks
• Research and experimentation
• Local AI with maximum freedom

Thank you for considering this request! The community would be very grateful for the complete vision-language setup. 🙏