W4A8KV4 Quantization Summary and Best Practices
Comprehensive summary of W4A8KV4 quantization techniques, covering KV4 and W4A8 optimization methods with practical recommendations.
Model Optimization
Low-Bit MoE Quantization for Large Language Models
Comprehensive guide to quantizing large MoE models like DeepSeek-V3/R1, covering techniques for efficient memory usage and inference optimization.
Speculative Sampling for Faster LLM Inference
Deep dive into speculative sampling techniques for accelerating LLM inference through draft model prediction and rejection sampling.