DLRM Tutorial: Deep Learning Recommendation Model

DLRM Tutorial: Deep Learning Recommendation Model#

Sandbox Model — Experimental

This model is part of the CoreRec Sandbox — an experimental collection of algorithms for research and exploration. Sandbox models are not production-tested and may have incomplete implementations. For production-ready models, see the Production Models.

Introduction#

DLRM separates dense and sparse feature processing, using explicit pairwise dot product interactions between embeddings.

Paper: Naumov et al. 2019 - Deep Learning Recommendation Model (Facebook)

How DLRM Works#

Architecture#

Parallel Dense/Sparse Processing:

  1. Bottom MLP: Process continuous features

  2. Embedding Layers: Parallel lookup for categorical features

  3. Explicit Interactions: Dot products between all embedding pairs

  4. Top MLP: Process concatenated features + interactions

Key Innovation: Scalable architecture for production (Facebook, Pinterest)

Architecture:

Dense → Bottom MLP →                        [Dot Products] → Concat → Top MLP → Pred
Sparse → Embed×K → /

Mathematical Foundation#

Dense Processing:

z_dense = MLP_bottom(x_dense)

Embedding Interactions (all pairs):

I = {<e_i, e_j> : for all i < j}
where <·,·> is dot product

Concatenation:

z = [z_dense; e_1; e_2; ...; e_K; I]

Final Prediction:

ŷ = σ(MLP_top(z))

Reference Implementation#

Implementation Note

The code examples below are reference implementations for learning purposes. This is a sandbox model and has not been production-tested. Verify behavior thoroughly before using in production.

Step 1: Import and Load Data#

from corerec.sandbox.collaborative_full.nn_base.DLRM_base import DLRM_base as DLRM
from cr_learn import ml_1m
from sklearn.model_selection import train_test_split
import numpy as np

# Load dataset with CORRECT API
data = ml_1m.load()  # Returns dict with 'ratings', 'users', 'movies'
ratings_df = data['ratings']  # DataFrame with user_id, movie_id, rating, timestamp

print(f"Loaded {len(ratings_df)} ratings")

# Split data
train_df, test_df = train_test_split(ratings_df, test_size=0.2, random_state=42)

# Extract arrays for model
train_users = train_df['user_id'].values
train_items = train_df['movie_id'].values
train_ratings = train_df['rating'].values

test_users = test_df['user_id'].values
test_items = test_df['movie_id'].values
test_ratings = test_df['rating'].values

Step 2: Initialize Model#

model = DLRM(
    name="DLRM_Model",
    bottom_mlp=[512, 256, 64],
    top_mlp=[512, 512, 256, 1],
    embedding_dim=64,
    epochs=20,
    batch_size=256,
    learning_rate=0.001,
    verbose=True
)

print(f"Initialized {model.name}")

Step 3: Train#

model.fit(
    user_ids=train_users,
    item_ids=train_items,
    ratings=train_ratings
)

print("Training complete!")

Step 4: Predict#

# Single prediction
score = model.predict(user_id=1, item_id=100)
print(f"Predicted score: {score:.3f}")

# Batch predictions
test_predictions = model.batch_predict(list(zip(test_users[:100], test_items[:100])))

Step 5: Recommend#

# Get top-10 recommendations for user
user_id = 1
recommendations = model.recommend(
    user_id=user_id,
    top_k=10
)

print(f"Top-10 recommendations for User {user_id}:")
for rank, item_id in enumerate(recommendations, 1):
    print(f"  {rank}. Item {item_id}")

Step 6: Evaluate#

from sklearn.metrics import mean_squared_error
import numpy as np

# Predict all test ratings
test_pred = [model.predict(u, i) for u, i in zip(test_users, test_items)]
rmse = np.sqrt(mean_squared_error(test_ratings, test_pred))
print(f"Test RMSE: {rmse:.4f}")

Step 7: Save & Load#

# Save model
model.save('artifacts/dlrm_model')

# Load model
loaded = DLRM.load('artifacts/dlrm_model')
test_score = loaded.predict(1, 100)
print(f"Loaded model prediction: {test_score:.3f}")

Key Takeaways#

When to Use DLRM#

✅ Ideal For:

  • Large-scale CTR prediction (billions of users)

  • Production systems (Facebook, Pinterest scale)

  • Mix of dense & sparse features

  • Parallelizable infrastructure

  • High-throughput serving

❌ Not For:

  • Small datasets (<100K)

  • Pure collaborative filtering

  • Sequential patterns

  • Limited compute/memory

Best Practices#

  1. Bottom MLP: [512, 256, 64]

  2. Top MLP: [512, 512, 256, 1]

  3. Embedding Dim: 16-128 (by cardinality)

  4. Batch Size: 2048-8192 (very large!)

  5. Mixed Precision: Use FP16 for speed

  6. Parallelization: Embeddings in parallel

  7. Caching: Cache popular item embeddings

  8. Hardware: Multi-GPU or TPU

Scaling & Production Considerations#

If you plan to move this sandbox model toward production use, consider:

  • Data Volume: Test with realistic dataset sizes; sandbox implementations may not be optimized for large-scale data.

  • Distributed Training: For datasets exceeding single-machine memory, consider wrapping the model in distributed training frameworks (e.g., PyTorch DDP, Horovod).

  • Hyperparameter Tuning: Use systematic search (Optuna, Ray Tune) rather than manual tuning. The defaults in sandbox models are not tuned for any specific domain.

  • Serving Latency: Profile inference time. Some architectures (e.g., attention-based) may need quantization or ONNX export for low-latency serving.

  • Monitoring: Implement A/B testing and online metrics (CTR, engagement, diversity) before replacing an existing system.

  • Validation: Write comprehensive unit tests and integration tests before deploying. Sandbox models do not have CI coverage.

Further Reading#

  • Paper: Naumov et al. 2019 - Deep Learning Recommendation Model (Facebook)

  • GitHub Repository

Contents