Skip to content
Other

WeVisOne-VL-GGML

GGUF-экспорты 2B и 4B для локального инференса

1 min readupdated

Репозиторий Asher-1 с GGUF-версиями WeVisOne-VL: моделями image-to-text на базе tencent/WeVisOne-VL-2B и tencent/WeVisOne-VL-4B. Экспорты предназначены для efficient local inference с llama.cpp и содержат VLM, мультимодальный проектор и mask decoder для сегментации.

Ключевые особенности

  • Два размера: 2B и 4B, файлы организованы по размеру модели.
  • VLM-варианты в FP16, Q8_0 и Q4_K_M; для 2B Q4_K_M весит 1.20 GB, для 4B — 2.53 GB.
  • Отдельные mmproj-файлы FP16: 0.76 GB для 2B и 0.78 GB для 4B.
  • Mask decoder есть в FP32 и FP16; Q4_K для decoder намеренно не предлагается.
  • Архитектура VLM-вариантов — qwen3vl; decoder использует wevisone_mask.

Пример

./llama-cli \
  -m wevisone-vlm-2b-q8_0.gguf \
  --mmproj mmproj-wevisone-vlm-2b-f16.gguf \
  --image your_image.jpg \
  -p "Describe this image."
ПодробнееFP16 VLM: 3.80 GB для 2B и 8.23 GB для 4B; Q8_0 — 2.02 и 4.38 GB.
Q8_0 заявлен как отдельный quantized tier с ratio ≤ 2.5 noise band; Q4_K_M — quality tier, требует проверки на parity suite.
Decoder FP32 — единственный parity-gated tier; FP16 может давать drift по MAD.
Требования к VRAM в карточке не указаны; mask decoder запускается через C++ runtime WeVisOne-VL.
Discuss Telegram 1 199

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Asher-1/WeVisOne-VL-GGML
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Asher-1/WeVisOne-VL-GGML --local-dir ~/models/WeVisOne-VL-GGML
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to WeVisOne-VL-GGML
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of WeVisOne-VL-GGML (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

Asher-1/WeVisOne-VL-GGML
Author
Asher-1
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

1991

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments