Labsco
firecrawl logo

knowledge-distillation

✓ Official★ 11

by firecrawl · part of firecrawl/ai-research-skills

Compress large language models using knowledge distillation from teacher to student models. Use when deploying smaller models with retained performance,…

🔥🔥🔥FreeQuick setup
#scraping#git
🧩 One of 7 skills in the firecrawl/ai-research-skills package — works on its own, and pairs well with its siblings.

Compress large language models using knowledge distillation from teacher to student models. Use when deploying smaller models with retained performance,…

Inspect the full instructions your agent will receiveExpand

This is the exact playbook injected into your agent when the skill activates — shown here so you can audit it before installing. You don't need to read it to use the skill.

name: knowledge-distillation description: Compress large language models using knowledge distillation from teacher to student models. Use when deploying smaller models with retained performance, transferring GPT-4 capabilities to open-source models, or reducing inference costs. Covers temperature scaling, soft targets, reverse KLD, logit distillation, and MiniLLM training strategies. version: 1.0.0 author: Orchestra Research license: MIT tags: [Emerging Techniques, Knowledge Distillation, Model Compression, Teacher-Student, MiniLLM, Reverse KLD, Soft Targets, Temperature Scaling, Logit Distillation, Model Transfer] dependencies: [transformers, torch, datasets]

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)

Core Concepts

1. Temperature Scaling

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

Copy & paste — that's it
# 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

Copy & paste — that's it
# 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

Copy & paste — that's it
# 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

Copy & paste — that's it
# 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

Copy & paste — that's it
# 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

Copy & paste — that's it
# 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

Best Practices

1. Hyperparameter Selection

Copy & paste — that's it
# 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

Copy & paste — that's it
# 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

Copy & paste — that's it
# 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

Copy & paste — that's it
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