Neural Magic Courses

Deploy LLMs More Efficiently with vLLM and Neural Magic

Explore vLLM's efficiency for LLM deployment and Neural Magic's enterprise solutions for cost-effective, scalable AI model services.

• YouTube
• 33 minutes
• On-Demand
• Free Video

Deploy Fast and Accurate YOLOv8 Object Detection Models on CPUs

Optimize YOLOv8 models for 10x smaller, 8x faster performance on CPUs. Learn sparsification techniques, quantization, and deployment strategies for efficient computer vision applications.

• YouTube
• 48 minutes
• On-Demand
• Free Video

Unlock Faster and More Efficient LLMs with SparseGPT - Neural Magic

Optimize LLMs with SparseGPT: prune and quantize models for efficient CPU deployment at GPU speeds. Explore one-shot compression, benchmarks, and practical implementation.

• YouTube
• 42 minutes
• On-Demand
• Free Video

Pruning and Quantizing ML Models With One Shot Without Retraining

Adapt advanced pruning and quantization methods for ML models, achieving 4X speedup with 99% accuracy in minutes, using Computer Vision and NLP examples.

• YouTube
• 53 minutes
• On-Demand
• Free Video

Applying Second-Order Pruning Algorithms for SOTA Model Compression

Explore second-order pruning algorithms for model compression, achieving higher sparsity while maintaining accuracy. Learn to apply these techniques to your ML projects for improved efficiency.

• YouTube
• 42 minutes
• On-Demand
• Free Video

Sparse Training of Neural Networks Using AC/DC

Explore AC/DC, a novel sparse training algorithm for DNNs. Learn its benefits, implementation, and deployment for improved performance and accuracy in deep learning.

• YouTube
• 42 minutes
• On-Demand
• Free Video

How Well Do Sparse Models Transfer? - Exploring Transfer Performance in Computer Vision and NLP

Explore sparse models' transfer performance and efficiency gains in image and language tasks, outperforming dense models even at high sparsities.

• YouTube
• 50 minutes
• On-Demand
• Free Video

The State of LLM Compression — From Research to Production

Explore the evolving landscape of LLM compression techniques, from cutting-edge research to practical implementation, examining quantization and sparsity tradeoffs for optimizing generative AI systems in production environments.

• YouTube
• 1 hour
• On-Demand
• Free Video

Synthetic Data Generation via SDG-Hub - Random Samples

Explore SDG Hub, an open-source toolkit for customizing language models with synthetic data, covering its components, strategies for teacher model selection, and real-world applications through demonstrations and examples.

• YouTube
• 55 minutes
• On-Demand
• Free Video

Continual Post-Training for Large Language Models

Explore continual learning for LLMs through a practical method that enables fine-tuning without compromising existing capabilities, using low-rank subspace constraints to preserve knowledge while adapting to new tasks.

• YouTube
• 1 hour 3 minutes
• On-Demand
• Free Video

Towards Combinatorial Interpretability of Neural Computation

Explore a novel combinatorial approach to neural network interpretability through the Feature Channel Coding Hypothesis, revealing how networks compute Boolean expressions and the natural limitations of code interference.

• YouTube
• 1 hour 1 minute
• On-Demand
• Free Video