Оптимизация инференса LLM on/theory/inference-optimization/Recent content in Оптимизация инференса LLM onHugo -- gohugo.iorurakhmankulovbulat@gmail.comrakhmankulovbulat@gmail.com© 2026Sat, 28 Feb 2026 00:00:00 +0000Ключевые метрики LLM инференса/theory/inference-optimization/llm-inference-metrics/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/llm-inference-metrics/Бенчмарки LLM/theory/inference-optimization/llm-performance-benchmarks/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/llm-performance-benchmarks/Статический, динамический и непрерывный батчинг/theory/inference-optimization/static-dynamic-continuous-batching/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/static-dynamic-continuous-batching/FlashAttention/theory/inference-optimization/flashattention/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/flashattention/PagedAttention/theory/inference-optimization/pagedattention/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/pagedattention/Speculative decoding/theory/inference-optimization/speculative-decoding/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/speculative-decoding/Дисагрегация prefill и decode/theory/inference-optimization/prefill-decode-disaggregation/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/prefill-decode-disaggregation/Кэширование префикса (Prefix caching)/theory/inference-optimization/prefix-caching/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/prefix-caching/Маршрутизация с учётом префикса/theory/inference-optimization/prefix-aware-routing/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/prefix-aware-routing/Балансировка нагрузки с учётом KV-кэша/theory/inference-optimization/kv-cache-utilization-aware-load-balancing/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/kv-cache-utilization-aware-load-balancing/Выгрузка KV-кэша (KV cache offloading)/theory/inference-optimization/kv-cache-offloading/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/kv-cache-offloading/Data, tensor, pipeline, expert и гибридный параллелизм/theory/inference-optimization/data-tensor-pipeline-expert-hybrid-parallelism/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/data-tensor-pipeline-expert-hybrid-parallelism/Оффлайн пакетный инференс/theory/inference-optimization/offline-batch-inference/Sat, 28 Feb 2026 00:00:00 +0000rakhmankulovbulat@gmail.com/theory/inference-optimization/offline-batch-inference/