New to Claude Skills? Learn how to install them →

davila7 on GitHub

Knowledge Distillation

Free

Compress large language models while retaining performance.

Get this skill

Free · Opens the source repo

What Knowledge Distillation does

Knowledge Distillation is a technique that enables the compression of large language models (LLMs) into smaller, more efficient versions without significantly sacrificing performance. This skill is particularly useful for developers and researchers who need to deploy smaller models in production environments while maintaining high accuracy. By transferring knowledge from a larger teacher model to a smaller student model, users can achieve significant reductions in model size, such as compressing a 70 billion parameter model to just 7 billion parameters while retaining over 90% of its performance.

This skill leverages various advanced techniques like temperature scaling, soft targets, reverse KLD, and logit distillation, which are crucial for effective model compression. For instance, temperature scaling helps in softening the output probability distributions of the teacher model, allowing the student model to learn from the teacher's uncertainty. Additionally, the use of reverse KLD is highlighted as a more effective method for generative model distillation, enabling the student to cover all modes of the teacher's output rather than just matching its mean.

Knowledge Distillation is ideal for scenarios where there is a need to transfer capabilities from proprietary models, such as GPT-4, to open-source models like LLaMA or Mistral. It also aids in reducing inference costs by deploying smaller models that are faster and require less computational power. This skill is particularly beneficial for creating specialized models that can handle domain-specific tasks by distilling knowledge from larger models.

Overall, this skill is a valuable tool for machine learning practitioners looking to optimize their model deployment strategies and improve the efficiency of their AI applications without compromising on performance.

When to use it

Use this skill when you need to compress large models for deployment, transfer capabilities from larger models to smaller ones, or reduce inference costs in your applications.

When not to use it

This skill may not be suitable for scenarios where the original model's performance is critical and cannot be compromised, or when working with models that are already optimized for size and performance.

What you can build with it

Deploying Smaller Models

Use this skill to compress a large language model, like GPT-4, into a smaller version for deployment in applications where computational resources are limited.

Transferring Model Capabilities

Leverage this skill to transfer capabilities from a proprietary model to an open-source model, enabling broader access to advanced AI functionalities.

Creating Domain-Specific Models

Utilize knowledge distillation to create specialized models that incorporate domain-specific knowledge, enhancing performance in targeted tasks.

How to install Knowledge Distillation

View source

1. Install with the skills CLI

npx skills add davila7/claude-code-templates/emerging-techniques-knowledge-distillation --agent claude-code

2. Or install it manually

Download the skill folder and drop it into ~/.claude/skills/ for all projects, or .claude/skills/ to scope it to one repo. Restart Claude Code so it picks up the new skill.

Anthropic's agentic coding CLI, and the reference implementation of Agent Skills. Drop a skill folder into ~/.claude/skills and Claude Code loads it automatically whenever a task matches the skill's description. Claude Code docs

Inside SKILL.md

Written by davila7

Knowledge Distillation: Compressing LLMs

When to Use This Skill

Use Knowledge Distillation when you need to:

  • Compress models from 70B → 7B while retaining 90%+ performance
  • Transfer capabilities from proprietary models (GPT-4) to open-source (LLaMA, Mistral)
  • Reduce inference costs by deploying smaller student models
  • Create specialized models by distilling domain-specific knowledge
  • Improve small models using synthetic data from large teachers

Key Techniques: Temperature scaling, soft targets, reverse KLD (MiniLLM), logit distillation, response distillation

Papers: Hinton et al. 2015 (arXiv 1503.02531), MiniLLM (arXiv 2306.08543), KD Survey (arXiv 2402.13116)

Installation

# Standard transformers
pip install transformers datasets accelerate

# For training
pip install torch deepspeed wandb

# Optional: MiniLLM implementation
git clone https://github.com/microsoft/LMOps
cd LMOps/minillm
pip install -e .

Quick Start

Basic Knowledge Distillation

import torch
import torch.nn.functional as F
from transformers import AutoModelForCausalLM, AutoTokenizer, Trainer, TrainingArguments

# 1. Load teacher (large) and student (small) models
teacher = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-70b-hf",  # Large teacher
    torch_dtype=torch.float16,
    device_map="auto"
)

student = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-hf",  # Small student
    torch_dtype=torch.float16,
    device_map="cuda:0"
)

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

# 2. Define distillation loss
def distillation_loss(student_logits, teacher_logits, labels, temperature=2.0, alpha=0.5):
    """
    Combine hard loss (cross-entropy) with soft loss (KL divergence).

    Args:
        temperature: Softens probability distributions (higher = softer)
        alpha: Weight for distillation loss (1-alpha for hard loss)
    """
    # Hard loss: Standard cross-entropy with true labels
    hard_loss = F.cross_entropy(student_logits.view(-1, student_logits.size(-1)), labels.view(-1))

    # Soft loss: KL divergence between student and teacher
    soft_targets = F.softmax(teacher_logits / temperature, dim=-1)
    soft_student = F.log_softmax(student_logits / temperature, dim=-1)
    soft_loss = F.kl_div(soft_student, soft_targets, reduction='batchmean') * (temperature ** 2)

    # Combined loss
    return alpha * soft_loss + (1 - alpha) * hard_loss

# 3. Training loop
for batch in dataloader:
    # Teacher forward (no grad)
    with torch.no_grad():
        teacher_outputs = teacher(**batch)
        teacher_logits = teacher_outputs.logits

    # Student forward
    student_outputs = student(**batch)
    student_logits = student_outputs.logits

    # Compute distillation loss
    loss = distillation_loss(
        student_logits,
        teacher_logits,
        batch['labels'],
        temperature=2.0,
        alpha=0.7  # 70% soft, 30% hard
    )

    # Backward and optimize
    loss.backward()
    optimizer.step()
    optimizer.zero_grad()

MiniLLM (Reverse KLD)

Source: arXiv 2306.08543 (2024)

Innovation: Use reverse KLD instead of forward KLD for better generative model distillation.

def reverse_kl_loss(student_logits, teacher_logits, temperature=1.0):
    """
    Reverse KL divergence: KL(Teacher || Student)
    Better for generative models than forward KL.
    """
    # Teacher distribution (target)
    p_teacher = F.softmax(teacher_logits / temperature, dim=-1)

    # Student distribution (model)
    log_p_student = F.log_softmax(student_logits / temperature, dim=-1)

    # Reverse KL: Sum over teacher, student learns to cover teacher's modes
    reverse_kl = -(p_teacher * log_p_student).sum(dim=-1).mean()

    return reverse_kl * (temperature ** 2)

# Training with MiniLLM
for batch in dataloader:
    with torch.no_grad():
        teacher_logits = teacher(**batch).logits

    student_logits = student(**batch).logits

    # Reverse KLD (better for generation)
    loss = reverse_kl_loss(student_logits, teacher_logits, temperature=1.0)

    loss.backward()
    optimizer.step()

Why reverse KL?

  • Forward KL (standard): Student learns to match teacher's mean
  • Reverse KL (MiniLLM): Student learns to cover all teacher's modes
  • Better for diverse text generation

Response Distillation

# Generate synthetic data from teacher, train student to imitate

# 1. Generate synthetic responses from teacher
prompts = ["Explain AI:", "What is ML?", "Define NLP:"]

teacher_responses = []
for prompt in prompts:
    inputs = tokenizer(prompt, return_tensors='pt').to(teacher.device)
    outputs = teacher.generate(**inputs, max_new_tokens=256, do_sample=True, temperature=0.7)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    teacher_responses.append(response)

# 2. Train student on teacher's responses (standard fine-tuning)
train_dataset = [
    {"text": f"{prompt}\n{response}"}
    for prompt, response in zip(prompts, teacher_responses)
]

# 3. Fine-tune student
trainer = Trainer(
    model=student,
    args=TrainingArguments(output_dir="./student", num_train_epochs=3, learning_rate=2e-5),
    train_dataset=train_dataset,
)
trainer.train()

Core Concepts

1. Temperature Scaling

Purpose: Soften probability distributions to expose teacher's uncertainty.

# Low temperature (T=1): Sharp distribution
logits = [3.0, 2.0, 1.0]
probs_T1 = softmax(logits / 1.0)  # [0.67, 0.24, 0.09]

# High temperature (T=4): Soft distribution
probs_T4 = softmax(logits / 4.0)  # [0.42, 0.34, 0.24]

# Higher T reveals more information about relative rankings

Rule: Use T=2-5 for distillation (2 is common default).

2. Loss Function Components

# Total loss = alpha * soft_loss + (1 - alpha) * hard_loss

# Soft loss: Learn from teacher's knowledge
soft_loss = KL(student || teacher)

# Hard loss: Learn from ground truth labels
hard_loss = CrossEntropy(student_output, true_labels)

# Typical values:
alpha = 0.5  # Balanced
alpha = 0.7  # More emphasis on teacher
alpha = 0.3  # More emphasis on labels

3. Forward vs Reverse KLD

# Forward KL: KL(Student || Teacher)
# - Student matches teacher's average behavior
# - Mode-seeking: Student focuses on teacher's highest probability modes
# - Good for classification

# Reverse KL: KL(Teacher || Student)
# - Student covers all of teacher's behaviors
# - Mode-covering: Student learns diverse behaviors
# - Good for generation (MiniLLM)

Training Strategies

Strategy 1: Logit Distillation

# Train student to match teacher's logits directly

def logit_distillation_trainer(student, teacher, dataloader, temperature=2.0):
    optimizer = torch.optim.AdamW(student.parameters(), lr=2e-5)

    for epoch in range(3):
        for batch in dataloader:
            # Get logits
            with torch.no_grad():
                teacher_logits = teacher(**batch).logits

            student_logits = student(**batch).logits

            # MSE on logits (alternative to KLD)
            loss = F.mse_loss(student_logits, teacher_logits)

            # Or use KLD
            # loss = F.kl_div(
            #     F.log_softmax(student_logits/temperature, dim=-1),
            #     F.softmax(teacher_logits/temperature, dim=-1),
            #     reduction='batchmean'
            # ) * (temperature ** 2)

            loss.backward()
            optimizer.step()
            optimizer.zero_grad()

    return student

Strategy 2: Two-Stage Distillation

# Stage 1: Distill from teacher
student = distill(teacher, student, epochs=5)

# Stage 2: Fine-tune on task-specific data
student = fine_tune(student, task_data, epochs=3)

# Results in better task performance than single-stage

Strategy 3: Multi-Teacher Distillation

# Learn from multiple expert teachers

def multi_teacher_distillation(student, teachers, batch):
    """Distill from ensemble of teachers."""
    teacher_logits_list = []

    # Get logits from all teachers
    with torch.no_grad():
        for teacher in teachers:
            logits = teacher(**batch).logits
            teacher_logits_list.append(logits)

    # Average teacher predictions
    avg_teacher_logits = torch.stack(teacher_logits_list).mean(dim=0)

    # Student learns from ensemble
    student_logits = student(**batch).logits
    loss = F.kl_div(
        F.log_softmax(student_logits, dim=-1),
        F.softmax(avg_teacher_logits, dim=-1),
        reduction='batchmean'
    )

    return loss

Production Deployment

Complete Training Script

from transformers import Trainer, TrainingArguments, DataCollatorForLanguageModeling

def train_distilled_model(
    teacher_name="meta-llama/Llama-2-70b-hf",
    student_name="meta-llama/Llama-2-7b-hf",
    output_dir="./distilled-llama-7b",
    temperature=2.0,
    alpha=0.7,
):
    # Load models
    teacher = AutoModelForCausalLM.from_pretrained(teacher_name, torch_dtype=torch.float16, device_map="auto")
    student = AutoModelForCausalLM.from_pretrained(student_name, torch_dtype=torch.float16)
    tokenizer = AutoTokenizer.from_pretrained(teacher_name)

    # Custom trainer with distillation
    class DistillationTrainer(Trainer):
        def compute_loss(self, model, inputs, return_outputs=False):
            # Student forward
            outputs_student = model(**inputs)
            student_logits = outputs_student.logits

            # Teacher forward (no grad)
            with torch.no_grad():
                outputs_teacher = teacher(**inputs)
                teacher_logits = outputs_teacher.logits

            # Distillation loss
            soft_targets = F.softmax(teacher_logits / temperature, dim=-1)
            soft_student = F.log_softmax(student_logits / temperature, dim=-1)
            soft_loss = F.kl_div(soft_student, soft_targets, reduction='batchmean') * (temperature ** 2)

            # Hard loss
            hard_loss = outputs_student.loss

            # Combined
            loss = alpha * soft_loss + (1 - alpha) * hard_loss

            return (loss, outputs_student) if return_outputs else loss

    # Training arguments
    training_args = TrainingArguments(
        output_dir=output_dir,
        num_train_epochs=3,
        per_device_train_batch_size=4,
        gradient_accumulation_steps=8,
        learning_rate=2e-5,
        warmup_steps=500,
        logging_steps=100,
        save_steps=1000,
        bf16=True,
        gradient_checkpointing=True,
    )

    # Train
    trainer = DistillationTrainer(
        model=student,
        args=training_args,
        train_dataset=train_dataset,
        data_collator=DataCollatorForLanguageModeling(tokenizer, mlm=False),
    )

    trainer.train()
    student.save_pretrained(output_dir)
    tokenizer.save_pretrained(output_dir)

# Usage
train_distilled_model(
    teacher_name="meta-llama/Llama-2-70b-hf",
    student_name="meta-llama/Llama-2-7b-hf",
    temperature=2.0,
    alpha=0.7
)

Best Practices

1. Hyperparameter Selection

# Temperature
T = 1.0  # Sharp (less knowledge transfer)
T = 2.0  # Standard (good balance)
T = 5.0  # Soft (more knowledge transfer)

# Alpha (weight)
alpha = 0.5  # Balanced
alpha = 0.7  # Emphasize teacher knowledge
alpha = 0.9  # Strong distillation

# Rule: Higher T + higher alpha = stronger distillation

2. Model Size Ratio

# Good ratios (teacher/student)
70B / 7B = 10×    # Excellent
13B / 1B = 13×    # Good
7B / 1B = 7×      # Acceptable

# Avoid too large gap
70B / 1B = 70×    # Too large, ineffective

3. Data Quality

# Best: Use teacher-generated data + real data
train_data = {
    "teacher_generated": 70%,  # Diverse, high-quality
    "real_data": 30%            # Ground truth
}

# Avoid: Only real data (doesn't utilize teacher fully)

Evaluation

from transformers import pipeline

# Compare student vs teacher
teacher_pipe = pipeline("text-generation", model=teacher)
student_pipe = pipeline("text-generation", model=student)

prompts = ["Explain quantum computing:", "What is AI?"]

for prompt in prompts:
    teacher_out = teacher_pipe(prompt, max_new_tokens=100)
    student_out = student_pipe(prompt, max_new_tokens=100)

    print(f"Prompt: {prompt}")
    print(f"Teacher: {teacher_out[0]['generated_text']}")
    print(f"Student: {student_out[0]['generated_text']}")
    print(f"Match quality: {calculate_similarity(teacher_out, student_out):.2f}")

Resources

Frequently asked questions about Knowledge Distillation

Similar skills