玩模型开放模型 · 清晰来源 · 自主选择
返回模型目录
raspie /

gemma-4-12B-it-abliterated-uncensored-gemmanima-argon-0.2.0

HF 源站

raspie/gemma-4-12B-it-abliterated-uncensored-gemmanima-argon-0.2.0 is an uncensored open-source model hosted on Hugging Face.

EnglishLicense: other12Bgemmaanimabridgeadapterimage-generationanimelocal-runtimeko
README.md说明 / 使用方法

language:

  • en
  • ko tags:
  • gemma
  • anima
  • bridge
  • adapter
  • image-generation
  • anime
  • local-runtime license: other

Gemma 4 12B IT Abliterated Uncensored - GemmAnima Argon Bridge

Korean model card | GitHub app source

Prototype Notice

This is the minimal v0.2.0 Argon bridge release for GemmAnima.

It contains only the evaluated Argon 100k direct Gemma 4 12B hidden-state bridge checkpoint plus metadata and license notices. It does not contain Gemma base weights, Anima base weights, GemmAnima chat LoRAs, vision/tagger LoRAs, planner adapters, old bridge experiments, or rollback checkpoints.

This is a public prototype artifact. It is not a production image model, not a safety-rated assistant, and not a commercial-ready release.

Uploaded Payload

FileRole
hiddenstage_bridge/argon_12b_100k_bridge.ptArgon 100k direct Gemma 4 12B hidden-state to Anima-conditioning bridge
adapter_manifest_v0.2.0_argon.jsonSHA256 and byte-size manifest
asset_manifest.jsonMinimal local placement manifest for this bridge
model_sources.jsonMinimal source metadata for this bridge
LICENSE_NOTICES.mdComposite prototype license notice
README.md / HF_MODEL_CARD.ko.mdModel cards

Base Models Are Not Included

Use the original pages for upstream base weights:

ComponentSource
Gemma 4 12B GGUFggml-org/gemma-4-12B-it-GGUF
Anima diffusion/VAEcirclestone-labs/Anima

The GitHub app source owns full runtime setup and model download planning. This HF repository is only the Argon bridge artifact.

Evaluation Evidence

Local checks before this release:

CheckResult
Argon 100k bridge final validation MSE0.0019557795094442554
Cached smoke renderPass
OOD g/e pose render, 3 prompts, Qwen baseline comparisonPass

The OOD test used fresh Gemma 4 12B hidden states for new prompts and rendered Qwen baseline vs Argon bridge outputs side by side. Results were judged usable for an Argon prototype: pose and composition were preserved well enough for this release line, while exact text rendering and strict semantic compliance still need more evaluation.

Intended Use

This bridge is intended for the GemmAnima standalone app and related local experiments that connect Gemma 4 12B hidden states to Anima-compatible conditioning.

Do not use this file as a standalone model. It requires the GemmAnima runtime code, upstream Gemma hidden-state generation, and Anima image-generation assets.

License And Restrictions

This bridge does not relicense upstream models. Users must comply with:

  • Gemma/GGUF source license terms
  • Anima source license terms
  • NVIDIA Open Model License terms where applicable through Anima's upstream lineage
  • GemmAnima adapter/checkpoint notices
  • Dataset and tag-source restrictions relevant to their use

Because the current image path depends on Anima, treat this Argon prototype as non-commercial, non-production, limited-test material unless you have separately verified all upstream permissions.

HUGGING FACE

上游数据

HF 下载量0
HF 点赞量1

HF 下载量为源站最近 30 天统计,与本站点击分开。

SITE STATS

本站统计

下载点击—
请求数—
匿名去重—最近 7 天—

本站下载量记录按钮点击,不代表下载完成;短时连点会去重。

来源与版本

作者
raspie
最近更新
2026/06/06
最新创建
2026/06/06
HF 趋势
未知
数据快照
2026/09/16
版本
未知
许可证
other