---
license: mit
tags:
  - speculative-decoding
  - llm-inference
  - benchmark
  - quantization
  - gguf
  - llama.cpp
---

# Speculative Decoding Lab

Systematic, reproducible empirical evaluation of **modern speculative decoding**
(EAGLE-3, DFlash, DSpark vs. vanilla and model-free) on an **RTX 4060 Ti 16 GB**
with **Gemma 4 12B** and **Qwen3-8B** GGUF-quantized models: acceptance by
cognitive domain, quantization effects, and hardware break-even maps.

> Paper: [PDF](paper/paper.pdf) · [arXiv-ready source bundle](paper/arxiv-bundle.tar.gz) · [source (markdown)](paper/paper.md) · [paper page](https://huggingface.co/spaces/reyden009/speculative-decoding-paper)

## Hardware used

The study was developed and benchmarked on this machine, so all reported
numbers (speedups, acceptance, break-even) are tied to these specs. The code
itself is portable: it only needs llama.cpp (CUDA build) and enough VRAM for
the target model, and can be re-run on any other hardware.

| Component | Detail |
| --- | --- |
| GPU | NVIDIA RTX 4060 Ti **16 GB** (Ada Lovelace, compute 8.9) |
| CPU | Intel i5-12400 (6C/12T, 4.4 GHz) |
| RAM / Disk | 31 GB / ~185 GB free |
| OS | Ubuntu 24.04 · driver 595.84 |

Environment principle: **everything user-space, no sudo** (uv + CUDA runfile in `$HOME`).

## Stack

- **Python 3.12 + uv** — `pyproject.toml` + `uv.lock` (numpy, ruff, pyright, cmake, ninja).
- **CUDA toolkit 12.8.1** in `~/cuda` (runfile `--toolkitpath`). *Note: the `nvcc` binary is not on PyPI (the wheels only ship ptxas).*
- **llama.cpp master 0713275 (v22)** (CUDA build) in `~/llama.cpp` (built from source; no CUDA binaries are published for Linux).
- **GGUF models** in `models/` (not versioned; downloaded with `curl -C -`).

## Setup (once)

```bash
uv sync                        # create .venv with all dependencies
source scripts/spec-env.sh     # PATH + CUDA_HOME + env vars (idempotent)
```

## Quality (required before committing)

```bash
uv run ruff check .            # lint
uv run ruff format --check .   # formatting
uv run pyright                 # types
```

## Execution regime

Overnight sessions with **pause/resume**: `scripts/bench_accept.py --resume`
(incremental JSONL per prompt; server-based runner with α/τ), downloads with
`scripts/dl.py`, runs in tmux/nohup. Measurement campaigns are launched with
`scripts/run_final_chain.sh`, `scripts/run_curves_chain.sh` and
`scripts/run_ksweep_chain.sh`; final analysis with `scripts/analyze_final.py`.

## Structure

```
scripts/       environment, downloads, benchmark runner and analysis (bench_accept.py, analyze_final.py)
experiments/   runs/ (raw, not versioned) and prompts/ (JSONL subsamples)
models/        GGUFs (not versioned)
```

## For agents

Read [`AGENTS.md`](AGENTS.md) before working in this repo (conventions, git
rules, sessions and handoff).

## License

MIT — see [`LICENSE`](LICENSE).