language:
- en
- zh
license: mit
base_model: apetersson/DeepSeek-V4-Flash-Vision-Exp-Abliterated
tags: - solstice-ai
- deepseek
- deepseek-v4
- gguf
- llama.cpp
- ollama
- vision
- multimodal
- mmproj
- dspark
- speculative-decoding
- uncensored
- abliterated
pipeline_tag: image-text-to-text
DeepSeek-V4-Flash-Vision-UNCENSORED (Official GGUF Release)
Official Solstice-AI GGUF Release • Pure BF16 Multimodal Vision Projector (`mmproj`) • Bundled DSpark Speculative Drafter
Original Architecture by DeepSeek AI • Uncensored Weights by apetersson • Quantization by Solstice-AI
Multimodal Vision Projector (mmproj)
In llama.cpp, multimodal models decouple language weights from vision encoding. This repository includes the official standalone pure bfloat16 Multimodal Projector:
mmproj-BF16.gguf(1.11 GB) — Native bfloat16 Vision Transformer (ViT) with temporal & spatial patch merging. Retains full FP32 dynamic range without FP16 underflow risks.
Quickstart with llama.cpp:
llama-server \
-m Q4_K_M/DeepSeek-V4-Flash-Vision-Uncensored-Q4_K_M-00001-of-00005.gguf \
--mmproj mmproj-BF16.gguf \
-c 131072 \
--port 8080
Speculative Decoding Acceleration
This release bundles pre-aligned speculative draft weights in speculative/:
speculative/DSpark-drafter-vision-exp.gguf
Run speculative decoding via llama-cli:
llama-cli \
-m Q4_K_M/DeepSeek-V4-Flash-Vision-Uncensored-Q4_K_M-00001-of-00005.gguf \
-md speculative/DSpark-drafter-vision-exp.gguf \
--mmproj mmproj-BF16.gguf \
-p "Describe this architecture diagram in detail." \
--image ./diagram.png