tags:
- bonfyre-fpq
- native-format
- compressed
- gemma-3
- abliterated
license: gemma
base_model: mlabonne/gemma-3-1b-it-abliterated
Gemma-3-1B-IT Abliterated — BonfyreFPQ Native Format
This is mlabonne/gemma-3-1b-it-abliterated compressed into the BonfyreFPQ v4 native format.
Compression Summary
| File | Tensors | BF16 Source | FPQ Size | File Ratio |
|---|---|---|---|---|
| model.fpq | 340 | 1.9 GB | 112 MB | ~17.4× |
- FPQ version: v4
- Algorithm: COORD@3+QJL (3-bit Lloyd-Max quantization + Johnson-Lindenstrauss transform)
- Average bpw: 3.50
- FP32-equivalent ratio: 9.1×
Format
bonfyre-fpq decompress model.fpq model.safetensors
Source
BF16 safetensors from mlabonne/gemma-3-1b-it-abliterated (single shard, 1.9 GB).
Abliteration removes refusal behavior. Compressed with BonfyreFPQ v4 on macOS/NEON.