---
base_model: prism-ml/Bonsai-27B-gguf
license: apache-2.0
pipeline_tag: text-generation
language:
- en
tags:
- bonsai
- qwen3_5
- antidoom
- ftpo
- anti-repetition
- 1-bit
- q1_0
- gguf
- speculative-decoding
- dspark
- llama.cpp
- blackwell
---

# Bonsai-27B-antidoom-1bit-DSpark

A **[DSpark](https://github.com/PrismML-Eng/Bonsai-demo)-accelerated + [antidoom](https://github.com/Liquid4All/antidoom)
(FTPO anti-repetition)** build of [`prism-ml/Bonsai-27B`](https://huggingface.co/prism-ml/Bonsai-27B-gguf) —
the 1-bit (`Q1_0`, 1.125 bpw g128) quantization of the Qwen3.5-27B hybrid-attention model.

Runs in the **[PrismML-Eng/llama.cpp](https://github.com/PrismML-Eng/llama.cpp)** fork (CUDA, the
Q1_0 hybrid-attention kernels). Benchmarked on a single **RTX 5090 (32 GB, Blackwell sm_120)**.

## Contents
| File | What |
|---|---|
| `Bonsai-27B-antidoom-1bit-Q1_0.gguf` | antidoom-tuned 1-bit model (4.5 GB) |
| `Bonsai-27B-dspark-Q4_1.gguf` | DSpark speculative drafter (from prism-ml) |
| `Bonsai-27B-mmproj-Q8_0.gguf` | vision projector (from prism-ml) |

## Run with DSpark (fastest)
```bash
build/bin/llama-server \
  -m Bonsai-27B-antidoom-1bit-Q1_0.gguf \
  -md Bonsai-27B-dspark-Q4_1.gguf \
  --spec-type draft-dspark --spec-draft-n-max 4 \
  -ngl 999 -ngld 999 -fa on -c 16384 -np 1
```

## Benchmark (RTX 5090, greedy, bs=1, 256 tok)
| Config | tok/s | accept | note |
|---|---|---|---|
| **DSpark, code prompt** | **186.2** | 78.6% | **1.16× over native** |
| Native (llama-bench tg256) | 161.2 | — | reference |
| Native (llama-server decode) | 157.5 | — | |
| DSpark, prose prompt | 132.8 | 51% | low-accept workloads slower |

DSpark speedup is workload-dependent: on predictable content (code, high accept) it beats native;
the 1-bit target is so bandwidth-cheap (3.5 GB) that low-accept drafting overhead can dominate.
The 5090 exceeds prism-ml's published H100 native (104.8 t/s) — this 1-bit model is memory-bandwidth-bound.

## Antidoom (FTPO) pass — what was done
Bonsai-27B is unusually **doom-loop resistant** (it self-corrects repetition and stays coherent even
under forced long generation), so runaway-repetition pairs were surfaced by **priming**: contexts that
have already begun repeating a phrase, sent to the raw completion endpoint, where the model continues
the loop. Antidoom's own detector + chosen-token sampler then extracted **40 FTPO preference pairs**
(rejected = the loop-continuation token, chosen = coherent escapes). FTPO trained a LoRA
(r=32, q/k/v/o + gate/up/down, 12 epochs, lr 3e-5) on the FP16 unpacked model:

- **chosen-win (prefers the coherent escape over continuing the loop): 0.11 → 0.66** (early-stopped).
- LoRA merged (CPU) into FP16, re-quantized to `Q1_0`. Note the aggressive 1-bit format partially
  dilutes the LoRA's fine adjustments, so the anti-repetition effect is strongest at higher precision.

Output remains coherent at 1-bit. This is a demonstration of the antidoom method on a
1-bit hybrid-attention target, packaged with the shipped DSpark drafter.

## Provenance
Base weights, DSpark drafter and mmproj are from [prism-ml](https://huggingface.co/prism-ml)
(re-quantized after the antidoom LoRA merge for the model weights; drafter/mmproj copied unchanged).
Built with [antidoom](https://github.com/Liquid4All/antidoom) and the PrismML llama.cpp fork.