DanTagGen delta rev2 needs about 1.7 GB of memory at Q8_0 (a 0.4 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.3 GB (Q6_K) to 0.8 GB (F16).
DanTagGen delta rev2 – 400M param LLaMA model for Danbooru tag generation
DanTagGen delta rev2 is a 0.39 billion parameter LLaMA‑based model trained from scratch on 7.2 M Danbooru images. It outputs tags for images with a 2048‑token context window. The model is released under CC‑BY‑SA‑4.0 and is available in FP16, Q8_0, and Q6_K GGUF formats.
Good at
- Fast inference with llama.cpp
- High tag diversity
- Supports Danbooru style tagging
Limits
- Requires 2048‑token context
- No vision input, only text prompts
- CC‑BY‑SA‑4.0 license requires attribution
Best for
- Generating tags for image datasets
- Enhancing SD prompts with tag lists
- Batch tagging in automated pipelines
Who it’s for. Developers and artists who need quick, high‑quality Danbooru tags for image generation or dataset curation. The model is lightweight enough for local deployment with modest GPU memory.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q6_K | 0.3 GB | 1.0 GB | Fits | Get file → | |
| Q8_0 | 0.4 GB | 1.1 GB | Fits | Get file → | |
| F16 | 0.8 GB | 1.5 GB | Fits | Get file → |
For 8 GB VRAM use Q6_K (0.32 GB, ~1.6 GB RAM at 8k context). For 12 GB VRAM use Q8_0 (0.42 GB, ~1.7 GB RAM). If more memory is available, F16 (0.79 GB, ~2.1 GB RAM) gives full precision.
What GPU do you need for DanTagGen delta rev2?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is F16 (0.8 GB file, about 2.1 GB in use).
Which DanTagGen delta rev2 quant should you choose?
Q8_0 (0.4 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is Q6_K at 0.3 GB, with the quality trade-offs that come with fewer bits. Very close to the original quality for most uses, at a clearly larger size than 4-bit.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about DanTagGen delta rev2
How much VRAM does DanTagGen delta rev2 need?
About 1.7 GB at Q8_0 with an 8K context window. That is the 0.4 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run DanTagGen delta rev2 on an 8 GB GPU?
Yes, with F16: the 0.8 GB file needs about 2.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of DanTagGen delta rev2 GGUF?
It depends on the quant: Q6_K is 0.3 GB, Q8_0 is 0.4 GB, F16 is 0.8 GB.
What context length does DanTagGen delta rev2 support?
The GGUF header lists a trained context length of 2,048 tokens (2K). The KV cache grows with the context you use: about 0.8 GB at 8K and 0.2 GB at 2K.
Where can I download DanTagGen delta rev2 GGUF?
The files are on Hugging Face in the KBlueLeaf/DanTagGen-delta-rev2 repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does DanTagGen delta rev2 use?
Hugging Face lists the license as cc-by-sa-4.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run DanTagGen delta rev2 on 8 GB VRAM?
Yes, the Q6_K quantized model is 0.32 GB and requires about 1.6 GB RAM at 8k context, fitting comfortably on 8 GB VRAM.
Does this model support image input?
No, DanTagGen delta rev2 only accepts text prompts and outputs tags; it does not process images directly.
What license does the model use?
The model is released under CC‑BY‑SA‑4.0, requiring attribution and share‑alike for derivative works.