← back to catalog · registered 2026-08-22 13:56

jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2-nvfp4

jiangchengchengNLP Llama 51B second-order
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/jiangchengchengNLP%2FLlama-4-Scout-17B-16E-Instruct-abliterated-v2-nvfp4"
Response includes
  • classification m1
  • files 26
  • hub_downloads_all_time 76,868
  • author_summary 6 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · lifetime
77K
39 last 30d - cooling
Likes
3
Model age
12mo ago
created 2025-10-09
Downloads over time
Now76.9K→from26↑295,581%
028.2K56.4K84.6K26 on Oct 8, 202576.9K on Oct 11Oct '25Dec '25FebAprJunAugOct
Oct 8, 2025 → Oct 11 · 92 snapshots · spans 368 days

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Variants by this author 2 formats · 208 downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
apache-2.0
Tags
safetensors llama4 base_model:jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2 base_model:quantized:jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2 license:apache-2.0 8-bit compressed-tensors region:us

Related

Total size
64.9 GB
Files
26
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2025-11-29 17:03

Files by quantization

Auxiliary files 26 files 64.9 GB
model-00013-of-00014.safetensors 4.66 GB 58003b11 download
model-00005-of-00014.safetensors 4.66 GB fb5a69ef download
model-00009-of-00014.safetensors 4.66 GB 5fd55d9d download
model-00001-of-00014.safetensors 4.65 GB d513217d download
model-00006-of-00014.safetensors 4.65 GB 0e23e49d download
model-00010-of-00014.safetensors 4.65 GB cfe7537e download
model-00002-of-00014.safetensors 4.65 GB aafd4d56 download
model-00008-of-00014.safetensors 4.64 GB 9aea0fe3 download
model-00007-of-00014.safetensors 4.64 GB 6acfb061 download
model-00012-of-00014.safetensors 4.64 GB f6515d1c download
model-00011-of-00014.safetensors 4.64 GB 97fef26d download
model-00004-of-00014.safetensors 4.64 GB 234ae0ac download
model-00003-of-00014.safetensors 4.64 GB c491f8fb download
model-00014-of-00014.safetensors 4.47 GB f99f166a download
tokenizer.json 26.7 MB 172c9eb4 download
tokenizer.model 3.45 MB d0bdbaf5 download
model.safetensors.index.json 1.23 MB 66cc432b download
tokenizer_config.json 227 KB 7095ee1e download
config.json 26.7 KB a40ab8e3 download
chat_template.jinja 7.33 KB 11170e0b download
README.md 2.23 KB e566704c download
.gitattributes 1.53 KB 52373fe2 download
preprocessor_config.json 636 B 4eef7723 download
special_tokens_map.json 448 B ff46145e download
generation_config.json 255 B d0d3a798 download
processor_config.json 128 B 2f3cae49 download

README current version from Hugging Face


license: apache-2.0
base_model:

  • jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2

Llama-4-Scout-17B-16E-Instruct-abliterated-v2-nvfp4

🧠 模型简介

Llama-4-Scout-17B-16E-Instruct-abliterated-v2-nvfp4 是基于原模型
jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2
通过 llm-compressor 工具进行压缩与优化后的版本。

该版本采用 NVFP4 (NVIDIA FP4) 精度格式与 Mixture-of-Experts (MoE) 架构,
在保持生成质量的同时显著提升了推理速度与显存利用率。


⚙️ 模型规格

项目 内容
基础架构 Llama 4
参数规模 17B
专家数 16 Experts (MoE)
精度 NVFP4
优化工具 llm-compressor
许可证 Apache 2.0
原始模型 jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2

💻 推理环境支持

显卡型号 架构 是否支持
NVIDIA H100 Hopper ✅ 支持
NVIDIA B200 Blackwell ✅ 支持
NVIDIA RTX 5090 Blackwell (SM120) ❌ 暂不支持

当前版本的 NVFP4-MoE 算子 暂未在 sm120 架构(如 RTX 5090)上适配。


🚀 推理示例

vllm serve  jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2-nvfp4  --port 6006 --host 0.0.0.0 --max-model-len 20000 --served-model-name llama4   --gpu-memory-utilization 0.95  --tensor-parallel-size 4

⚙️ 推理参数

temperature = 0.7
top_p = 0.95

🧩 模型特点

  • ⚡ 使用 NVFP4 量化格式,显著减少显存占用。
  • 🔧 支持 **vLLM 推理框架。
  • 🧰 由 llm-compressor 自动优化算子、融合层、权重量化。

📄 许可证

本模型依据 Apache License 2.0 授权,可用于商业用途。


🙏 致谢

  • 原模型作者:@jiangchengchengNLP
  • 压缩与适配:llm-compressor 项目团队
  • 基础架构:Meta Llama 4

README history 6 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2025-10-11Update README.md (#1)ceeea0f2.2 KB
    Loading...
  2. 2025-10-09Update README.md882caab2.2 KB
    Loading...
  3. 2025-10-09Update README.mdfd097cf2.2 KB
    Loading...
  4. 2025-10-09Update README.md04de1f42.1 KB
    Loading...
  5. 2025-10-09Update README.md4a257762.5 KB
    Loading...
  6. 2025-10-09initial commitbefbaf228 B
    Loading...

Discussions 2 threads

  1. 2025-11-29PRDelete Junkmerged1 💬#2
    Loading...
  2. 2025-10-11PRUpdate README.mdmerged1 💬#1
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration