---
license: apache-2.0
base_model: jialinyyzz/humanizer
pipeline_tag: text-generation
library_name: transformers
tags:
- bf16
- humanizer
- full-precision
- transformers
- vllm
- xinference
---
# humanizer-bf16 (bf16)
Full-precision (bf16) PyTorch port of [jialinyyzz/humanizer](https://huggingface.co/jialinyyzz/humanizer),
for vLLM / TGI / Xinference / transformers users who don't want the GGUF quant.
- Format: `pytorch_model.bin` (bf16). safetensors conversion in progress.
- Architecture: identical to the original — drop-in for `AutoModelForCausalLM.from_pretrained("rty813/humanizer-bf16")`.
- Also loads in vLLM (`vllm serve rty813/humanizer-bf16`) and Xinference (search "humanizer").
Original model card:
---
license: apache-2.0
base_model: google/gemma-4-12B
base_model_relation: finetune
language:
- en
- zh
pipeline_tag: text-generation
library_name: gguf
tags:
- humanizer
- transformers
- safetensors
- text-rewriting
- rewriting
- paraphrase
- style-transfer
- gguf
- llama.cpp
- gemma4
---
# humanizer
**A 12B model that rewrites AI-written drafts (emails, essays, reports, forum posts; English and Chinese) so they read like a person wrote them.** It is trained to keep every number, unit, date, name and quote and to add nothing, and it runs locally.
**Quick start:** get the app for **[Mac (Apple silicon)](https://github.com/sgaofen/humanizer-local-model/releases/download/app-v0.3.2/Humanizer-0.3.2-macos-arm64.dmg)** or **[Windows](https://github.com/sgaofen/humanizer-local-model/releases/download/app-v0.3.2/Humanizer-0.3.2-windows-x64-setup.exe)**, open it and paste a draft; it works offline. Without the app: [llama.cpp and other runtimes](#usage-without-the-app), or [AGENTS.md](AGENTS.md) for coding agents.
## Before and after
Drafts from the held-out evaluation set. For each we generated 8 samples with the Q8_0 file (llama.cpp, the app's settings) and picked the one that reads best