gemma-4-31B-it-qat-q4_0-unquantized-uncensored-heretic-exl3-mul1

EXL3 (ExLlamaV3) quantizations of llmfan46/gemma-4-31B-it-qat-q4_0-unquantized-uncensored-heretic. All credit for the original model goes to the original authors.

📊 Available Quantizations & VRAM

The model weights are stored in separate branches. Please switch to a branch to download. Note: VRAM estimates include PyTorch context overhead (~0.8GB) and assume an unquantized FP16 KV cache.

Target BPW	Head BPW	Branch (Download Link)	WikiText-2 PPL (512 ctx)¹	2K ctx	4K ctx	8K ctx	16K ctx	32K ctx
3.5	h6	3.5bpw_h6	2335.9614	N/A	N/A	N/A	N/A	N/A
3.75	h6	3.75bpw_h6	2312.4667	N/A	N/A	N/A	N/A	N/A
4.0	h6	4.0bpw_h6	2097.6864	N/A	N/A	N/A	N/A	N/A
5.0	h6	5.0bpw_h6	1995.7812	N/A	N/A	N/A	N/A	N/A
6.0	h6	6.0bpw_h6	1995.3135	N/A	N/A	N/A	N/A	N/A
8.0	h8	8.0bpw_h8	1985.1562	N/A	N/A	N/A	N/A	N/A

¹ Evaluated against WikiText-2 with ExLlamaV3 using a strided 512-token context window (-c 512) in llama.cpp parity mode (-g). Lower is better.

(Higher BPW = higher quality, lower BPW = fits in less VRAM).

📥 How to Download

It's recommended to use the huggingface-cli to download specific branches. (Do not use git clone as it will download all branches!)

Ensure you have the CLI installed:

pip install -U "huggingface_hub[cli]"

Download a specific branch (e.g., 5.0bpw_h6):

# Example: Downloading the 5.0bpw_h6 branch
huggingface-cli download DBMe/gemma-4-31B-it-qat-q4_0-unquantized-uncensored-heretic-exl3-mul1 --revision 5.0bpw_h6 --local-dir gemma-4-31B-it-qat-q4_0-unquantized-uncensored-heretic-exl3-mul1-5.0bpw_h6

💻 Supported Engines

These models are highly optimized for modern GPUs and can be run using:

• TabbyAPI: A fast, OpenAI-compatible API server. (Set model_name to the local folder name you downloaded the branch into)
• Text-Generation-WebUI: A local web interface. (Select the exllamav3 loader)
• ExLlamaV3 (Native): Python library for custom integration.

📈 Perplexity Degradation Curve

(Lower is better)

⚙️ Advanced: Quantization Environment & Settings

🔬 Quantization Settings

• Codebook: mul1

• Output Scales: always

• Calibration Rows: 250

• Calibration Cols: 2048

• Calibration Dataset: ExLlamaV3 Default (Wiki/C4/Code)

• High Quality (HQ) Mode: False

• ExLlamaV3: 1.0.0 (Commit: cb7f2e9)

• Hardware: NVIDIA RTX PRO 6000 Blackwell Server Edition