home / skills / orchestra-research / ai-research-skills / speculative-decoding

This skill accelerates LLM inference using speculative decoding, Medusa heads, and lookahead techniques to boost speed and reduce latency.

npx playbooks add skill orchestra-research/ai-research-skills --skill speculative-decoding

Review the files below or copy the command above to add this skill to your agents.

Files (3)
SKILL.md
13.7 KB
---
name: speculative-decoding
description: Accelerate LLM inference using speculative decoding, Medusa multiple heads, and lookahead decoding techniques. Use when optimizing inference speed (1.5-3.6× speedup), reducing latency for real-time applications, or deploying models with limited compute. Covers draft models, tree-based attention, Jacobi iteration, parallel token generation, and production deployment strategies.
version: 1.0.0
author: Orchestra Research
license: MIT
tags: [Emerging Techniques, Speculative Decoding, Medusa, Lookahead Decoding, Fast Inference, Draft Models, Tree Attention, Parallel Generation, Latency Reduction, Inference Optimization]
dependencies: [transformers, torch]
---

# Speculative Decoding: Accelerating LLM Inference

## When to Use This Skill

Use Speculative Decoding when you need to:
- **Speed up inference** by 1.5-3.6× without quality loss
- **Reduce latency** for real-time applications (chatbots, code generation)
- **Optimize throughput** for high-volume serving
- **Deploy efficiently** on limited hardware
- **Generate faster** without changing model architecture

**Key Techniques**: Draft model speculative decoding, Medusa (multiple heads), Lookahead Decoding (Jacobi iteration)

**Papers**: Medusa (arXiv 2401.10774), Lookahead Decoding (ICML 2024), Speculative Decoding Survey (ACL 2024)

## Installation

```bash
# Standard speculative decoding (transformers)
pip install transformers accelerate

# Medusa (multiple decoding heads)
git clone https://github.com/FasterDecoding/Medusa
cd Medusa
pip install -e .

# Lookahead Decoding
git clone https://github.com/hao-ai-lab/LookaheadDecoding
cd LookaheadDecoding
pip install -e .

# Optional: vLLM with speculative decoding
pip install vllm
```

## Quick Start

### Basic Speculative Decoding (Draft Model)

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load target model (large, slow)
target_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-70b-hf",
    device_map="auto",
    torch_dtype=torch.float16
)

# Load draft model (small, fast)
draft_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-hf",
    device_map="auto",
    torch_dtype=torch.float16
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-70b-hf")

# Generate with speculative decoding
prompt = "Explain quantum computing in simple terms:"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")

# Transformers 4.36+ supports assisted generation
outputs = target_model.generate(
    **inputs,
    assistant_model=draft_model,  # Enable speculative decoding
    max_new_tokens=256,
    do_sample=True,
    temperature=0.7,
)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)
```

### Medusa (Multiple Decoding Heads)

```python
from medusa.model.medusa_model import MedusaModel

# Load Medusa-enhanced model
model = MedusaModel.from_pretrained(
    "FasterDecoding/medusa-vicuna-7b-v1.3",  # Pre-trained with Medusa heads
    torch_dtype=torch.float16,
    device_map="auto"
)

tokenizer = AutoTokenizer.from_pretrained("FasterDecoding/medusa-vicuna-7b-v1.3")

# Generate with Medusa (2-3× speedup)
prompt = "Write a Python function to calculate fibonacci numbers:"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")

outputs = model.medusa_generate(
    **inputs,
    max_new_tokens=256,
    temperature=0.7,
    posterior_threshold=0.09,  # Acceptance threshold
    posterior_alpha=0.3,       # Tree construction parameter
)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)
```

### Lookahead Decoding (Jacobi Iteration)

```python
from lookahead.lookahead_decoding import LookaheadDecoding

# Load model
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-hf",
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")

# Initialize lookahead decoding
lookahead = LookaheadDecoding(
    model=model,
    tokenizer=tokenizer,
    window_size=15,    # Lookahead window (W)
    ngram_size=5,      # N-gram size (N)
    guess_size=5       # Number of parallel guesses
)

# Generate (1.5-2.3× speedup)
prompt = "Implement quicksort in Python:"
output = lookahead.generate(prompt, max_new_tokens=256)
print(output)
```

## Core Concepts

### 1. Speculative Decoding (Draft Model)

**Idea**: Use small draft model to generate candidates, large target model to verify in parallel.

**Algorithm**:
1. Draft model generates K tokens speculatively
2. Target model evaluates all K tokens in parallel (single forward pass)
3. Accept tokens where draft and target agree
4. Reject first disagreement, continue from there

```python
def speculative_decode(target_model, draft_model, prompt, K=4):
    """Speculative decoding algorithm."""
    # 1. Generate K draft tokens
    draft_tokens = draft_model.generate(prompt, max_new_tokens=K)

    # 2. Target model evaluates all K tokens in one forward pass
    target_logits = target_model(draft_tokens)  # Parallel!

    # 3. Accept/reject based on probability match
    accepted = []
    for i in range(K):
        p_draft = softmax(draft_model.logits[i])
        p_target = softmax(target_logits[i])

        # Acceptance probability
        if random.random() < min(1, p_target[draft_tokens[i]] / p_draft[draft_tokens[i]]):
            accepted.append(draft_tokens[i])
        else:
            break  # Reject, resample from target

    return accepted
```

**Performance**:
- Speedup: 1.5-2× with good draft model
- Zero quality loss (mathematically equivalent to target model)
- Best when draft model is 5-10× smaller than target

### 2. Medusa (Multiple Decoding Heads)

**Source**: arXiv 2401.10774 (2024)

**Innovation**: Add multiple prediction heads to existing model, predict future tokens without separate draft model.

**Architecture**:
```
Input → Base LLM (frozen) → Hidden State
                                ├→ Head 1 (predicts token t+1)
                                ├→ Head 2 (predicts token t+2)
                                ├→ Head 3 (predicts token t+3)
                                └→ Head 4 (predicts token t+4)
```

**Training**:
- **Medusa-1**: Freeze base LLM, train only heads
  - 2.2× speedup, lossless
- **Medusa-2**: Fine-tune base LLM + heads together
  - 2.3-3.6× speedup, better quality

**Tree-based Attention**:
```python
# Medusa constructs tree of candidates
# Example: Predict 2 steps ahead with top-2 per step

#         Root
#        /    \
#      T1a    T1b  (Step 1: 2 candidates)
#     /  \    / \
#  T2a  T2b T2c T2d  (Step 2: 4 candidates total)

# Single forward pass evaluates entire tree!
```

**Advantages**:
- No separate draft model needed
- Minimal training (only heads)
- Compatible with any LLM

### 3. Lookahead Decoding (Jacobi Iteration)

**Source**: ICML 2024

**Core idea**: Reformulate autoregressive decoding as solving system of equations, solve in parallel using Jacobi iteration.

**Mathematical formulation**:
```
Traditional:  y_t = f(x, y_1, ..., y_{t-1})  (sequential)
Jacobi:       y_t^{(k+1)} = f(x, y_1^{(k)}, ..., y_{t-1}^{(k)})  (parallel)
```

**Two branches**:

1. **Lookahead Branch**: Generate n-grams in parallel
   - Window size W: How many steps to look ahead
   - N-gram size N: How many past tokens to use

2. **Verification Branch**: Verify promising n-grams
   - Match n-grams with generated tokens
   - Accept if first token matches

```python
class LookaheadDecoding:
    def __init__(self, model, window_size=15, ngram_size=5):
        self.model = model
        self.W = window_size  # Lookahead window
        self.N = ngram_size   # N-gram size

    def generate_step(self, tokens):
        # Lookahead branch: Generate W × N candidates
        candidates = {}
        for w in range(1, self.W + 1):
            for n in range(1, self.N + 1):
                # Generate n-gram starting at position w
                ngram = self.generate_ngram(tokens, start=w, length=n)
                candidates[(w, n)] = ngram

        # Verification branch: Find matching n-grams
        verified = []
        for ngram in candidates.values():
            if ngram[0] == tokens[-1]:  # First token matches last input
                if self.verify(tokens, ngram):
                    verified.append(ngram)

        # Accept longest verified n-gram
        return max(verified, key=len) if verified else [self.model.generate_next(tokens)]
```

**Performance**:
- Speedup: 1.5-2.3× (up to 3.6× for code generation)
- No draft model or training needed
- Works out-of-the-box with any model

## Method Comparison

| Method | Speedup | Training Needed | Draft Model | Quality Loss |
|--------|---------|-----------------|-------------|--------------|
| **Draft Model Speculative** | 1.5-2× | No | Yes (external) | None |
| **Medusa** | 2-3.6× | Minimal (heads only) | No (built-in heads) | None |
| **Lookahead** | 1.5-2.3× | None | No | None |
| **Naive Batching** | 1.2-1.5× | No | No | None |

## Advanced Patterns

### Training Medusa Heads

```python
from medusa.model.medusa_model import MedusaModel
from medusa.model.kv_cache import initialize_past_key_values
import torch.nn as nn

# 1. Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    "lmsys/vicuna-7b-v1.3",
    torch_dtype=torch.float16
)

# 2. Add Medusa heads
num_heads = 4
medusa_heads = nn.ModuleList([
    nn.Linear(base_model.config.hidden_size, base_model.config.vocab_size, bias=False)
    for _ in range(num_heads)
])

# 3. Training loop (freeze base model for Medusa-1)
for param in base_model.parameters():
    param.requires_grad = False  # Freeze base

optimizer = torch.optim.Adam(medusa_heads.parameters(), lr=1e-3)

for batch in dataloader:
    # Forward pass
    hidden_states = base_model(**batch, output_hidden_states=True).hidden_states[-1]

    # Predict future tokens with each head
    loss = 0
    for i, head in enumerate(medusa_heads):
        logits = head(hidden_states)
        # Target: tokens shifted by (i+1) positions
        target = batch['input_ids'][:, i+1:]
        loss += F.cross_entropy(logits[:, :-i-1], target)

    # Backward
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()
```

### Hybrid: Speculative + Medusa

```python
# Use Medusa as draft model for speculative decoding
draft_medusa = MedusaModel.from_pretrained("medusa-vicuna-7b")
target_model = AutoModelForCausalLM.from_pretrained("vicuna-33b")

# Draft generates multiple candidates with Medusa
draft_tokens = draft_medusa.medusa_generate(prompt, max_new_tokens=5)

# Target verifies in single forward pass
outputs = target_model.generate(
    prompt,
    assistant_model=draft_medusa,  # Use Medusa as draft
    max_new_tokens=256
)

# Combines benefits: Medusa speed + large model quality
```

### Optimal Draft Model Selection

```python
def select_draft_model(target_model_size, target):
    """Select optimal draft model for speculative decoding."""
    # Rule: Draft should be 5-10× smaller
    if target_model_size == "70B":
        return "7B"  # 10× smaller
    elif target_model_size == "33B":
        return "7B"  # 5× smaller
    elif target_model_size == "13B":
        return "1B"  # 13× smaller
    else:
        return None  # Target too small, use Medusa/Lookahead instead

# Example
draft = select_draft_model("70B", target_model)
# Returns "7B" → Use Llama-2-7b as draft for Llama-2-70b
```

## Best Practices

### 1. Choose the Right Method

```python
# New deployment → Medusa (best overall speedup, no draft model)
if deploying_new_model:
    use_method = "Medusa"

# Existing deployment with small model available → Draft speculative
elif have_small_version_of_model:
    use_method = "Draft Model Speculative"

# Want zero training/setup → Lookahead
elif want_plug_and_play:
    use_method = "Lookahead Decoding"
```

### 2. Hyperparameter Tuning

**Draft Model Speculative**:
```python
# K = number of speculative tokens
K = 4  # Good default
K = 2  # Conservative (higher acceptance)
K = 8  # Aggressive (lower acceptance, but more when accepted)

# Rule: Larger K → more speedup IF draft model is good
```

**Medusa**:
```python
# Posterior threshold (acceptance confidence)
posterior_threshold = 0.09  # Standard (from paper)
posterior_threshold = 0.05  # More conservative (slower, higher quality)
posterior_threshold = 0.15  # More aggressive (faster, may degrade quality)

# Tree depth (how many steps ahead)
medusa_choices = [[0], [0, 0], [0, 1], [0, 0, 0]]  # Depth 3 (standard)
```

**Lookahead**:
```python
# Window size W (lookahead distance)
# N-gram size N (context for generation)

# 7B model (more resources)
W, N = 15, 5

# 13B model (moderate)
W, N = 10, 5

# 33B+ model (limited resources)
W, N = 7, 5
```

### 3. Production Deployment

```python
# vLLM with speculative decoding
from vllm import LLM, SamplingParams

# Initialize with draft model
llm = LLM(
    model="meta-llama/Llama-2-70b-hf",
    speculative_model="meta-llama/Llama-2-7b-hf",  # Draft model
    num_speculative_tokens=5,
    use_v2_block_manager=True,
)

# Generate
prompts = ["Tell me about AI:", "Explain quantum physics:"]
sampling_params = SamplingParams(temperature=0.7, max_tokens=256)

outputs = llm.generate(prompts, sampling_params)
for output in outputs:
    print(output.outputs[0].text)
```

## Resources

- **Medusa Paper**: https://arxiv.org/abs/2401.10774
- **Medusa GitHub**: https://github.com/FasterDecoding/Medusa
- **Lookahead Decoding (ICML 2024)**: https://lmsys.org/blog/2023-11-21-lookahead-decoding/
- **Lookahead GitHub**: https://github.com/hao-ai-lab/LookaheadDecoding
- **Speculative Decoding Survey (ACL 2024)**: https://aclanthology.org/2024.findings-acl.456.pdf
- **Comprehensive Survey**: https://arxiv.org/abs/2401.07851

## See Also

- `references/draft_model.md` - Draft model selection and training
- `references/medusa.md` - Medusa architecture and training
- `references/lookahead.md` - Lookahead decoding implementation details


Overview

This skill accelerates LLM inference using speculative decoding, Medusa multiple heads, and lookahead (Jacobi) decoding to reduce latency and increase throughput. It targets real-time and resource-constrained deployments, delivering typical speedups of 1.5–3.6× while preserving model quality. The content covers draft-model workflows, Medusa head training, lookahead parameters, and production deployment patterns.

How this skill works

Speculative decoding uses a small draft model to propose token candidates and a large target model to verify them in parallel, accepting tokens when probabilities align. Medusa augments a frozen base LLM with multiple prediction heads that predict future tokens in one forward pass, enabling tree-based verification without a separate draft model. Lookahead decoding reformulates autoregressive decoding as a parallel Jacobi iteration that generates and verifies n-gram candidates across a sliding window.

When to use it

  • When you need 1.5–3.6× inference speedup without quality loss
  • To cut latency for chatbots, code generation, or interactive agents
  • When serving high-volume traffic on constrained hardware
  • If you want to deploy faster generation without changing core model architecture
  • When you prefer minimal training (Medusa heads) or plug-and-play methods (Lookahead)

Best practices

  • Select a draft model 5–10× smaller than the target for best speculative gains
  • Tune K (speculative tokens), posterior_threshold, and window/W/N parameters per model size
  • Use Medusa for new deployments when you can add heads; freeze base LLM initially (Medusa-1)
  • Apply Lookahead for zero-training, out-of-the-box speedups and simple integration
  • Benchmark end-to-end quality and latency with realistic prompts before production rollout

Example use cases

  • Real-time customer support chatbot requiring sub-second responses
  • Code generation service where throughput and latency both matter
  • Deploying large LLMs on limited GPU fleets to increase serving capacity
  • Hybrid setups: Medusa as draft for speculative decoding to combine speed and final-model quality
  • vLLM production pipelines using speculative_model + num_speculative_tokens for scale

FAQ

Will speculative decoding change output quality?

No. Proper speculative decoding with verification is mathematically equivalent to the target model and preserves quality when configured correctly.

When should I train Medusa heads versus using Lookahead?

Train Medusa heads if you control model training and want the best speedups (2–3.6×). Use Lookahead when you need plug-and-play acceleration with no training.