v2.6.4

What's Changed

• 🔥[BitNet] BitNet a4.8: 4-bit Activations for 1-bit LLMs by @DefTruth in #91

Full Changelog: v2.6.3...v2.6.4

v2.6.3

What's Changed

• 🔥[Fast Best-of-N] Fast Best-of-N Decoding via Speculative Rejection by @DefTruth in #89
• 🔥[Tensor Product] Acceleration of Tensor-Product Operations with Tensor Cores by @DefTruth in #90

Full Changelog: v2.6.2...v2.6.3

v2.6.2

What's Changed

• early exit of LLM inference by @boyi-liu in #85
• Add paper AdaKV by @FFY0 in #86
• Efficient Hybrid Inference for LLMs: Reward-Based Token Modelling with Selective Cloud Assistance by @aharshms in #87
• 🔥[FastAttention] FastAttention: Extend FlashAttention2 to NPUs and Low-resource GPUs for Efficient Inference by @DefTruth in #88

New Contributors

• @boyi-liu made their first contribution in #85
• @FFY0 made their first contribution in #86
• @aharshms made their first contribution in #87

Full Changelog: v2.6.1...v2.6.2

v2.6.1

What's Changed

• [From Author] Link CacheGen and CacheBlend to LMCache by @KuntaiDu in #80
• 🔥[LORC] Low-Rank Compression for LLMs KV Cache with a Progressive Compression Strategy by @DefTruth in #81
• Large Language Model Performance Benchmarking on Mobile Platforms: A Thorough Evaluation by @DefTruth in #82
• [LLM Inference] LARGE LANGUAGE MODEL INFERENCE ACCELERATION: A COMPREHENSIVE HARDWARE PERSPECTIVE by @DefTruth in #83
• 🔥[PARALLELSPEC] PARALLELSPEC: PARALLEL DRAFTER FOR EFFICIENT SPECULATIVE DECODING by @DefTruth in #84

New Contributors

• @KuntaiDu made their first contribution in #80

Full Changelog: v2.6...v2.6.1

v2.6

What's Changed

• 🔥[VPTQ] VPTQ: EXTREME LOW-BIT VECTOR POST-TRAINING QUANTIZATION FOR LARGE LANGUAGE MODELS by @DefTruth in #70
• fix typo by @DefTruth in #71
• 🔥🔥[INT-FLASHATTENTION] INT-FLASHATTENTION: ENABLING FLASH ATTENTION FOR INT8 QUANTIZATION by @DefTruth in #72
• [Low-bit] A Survey of Low-bit Large Language Models: Basics, Systems, and Algorithms by @DefTruth in #73
• 🔥🔥[HiFloat8] Ascend HiFloat8 Format for Deep Learning by @DefTruth in #74
• 🔥[AlignedKV] AlignedKV: Reducing Memory Access of KV-Cache with Precision-Aligned Quantization by @DefTruth in #75
• 🔥🔥[Tensor Cores] Efficient Arbitrary Precision Acceleration for Large Language Models on GPU Tensor Cores by @DefTruth in #76
• 🔥[KV-COMPRESS] PAGED KV-CACHE COMPRESSION WITH VARIABLE COMPRESSION RATES PER ATTENTION HEAD by @DefTruth in #77
• 🔥[LayerKV] Optimizing Large Language Model Serving with Layer-wise KV Cache Management by @DefTruth in #78
• Bump up to v2.6 by @DefTruth in #79

Full Changelog: v2.5...v2.6

v2.5

What's Changed

• 🔥[InstInfer] InstInfer: In-Storage Attention Offloading for Cost-Effective Long-Context LLM Inference by @DefTruth in #65
• Update codebase of paper "parallel speculative decoding with adaptive draft length" by @smart-lty in #66
• move RetrievalAttention -> long context by @DefTruth in #67
• 🔥🔥[CRITIPREFILL] CRITIPREFILL: A SEGMENT-WISE CRITICALITYBASED APPROACH FOR PREFILLING ACCELERATION IN LLMS by @DefTruth in #68
• Bump up to v2.5 by @DefTruth in #69

New Contributors

• @smart-lty made their first contribution in #66

Full Changelog: v2.4...v2.5

v2.4

What's Changed

• 🔥[RetrievalAttention] Accelerating Long-Context LLM Inference via Vector Retrieval by @DefTruth in #62
• 🔥[Inf-MLLM] Inf-MLLM: Efficient Streaming Inference of Multimodal Large Language Models on a Single GPU by @DefTruth in #63
• Bump up to v2.4 by @DefTruth in #64

Full Changelog: v2.3...v2.4

v2.3

What's Changed

• 🔥[CHESS] CHESS : Optimizing LLM Inference via Channel-Wise Thresholding and Selective Sparsification by @DefTruth in #59
• 🔥[SpMM] High Performance Unstructured SpMM Computation Using Tensor Cores by @DefTruth in #60
• Bump up to v2.3 by @DefTruth in #61

Full Changelog: v2.2...v2.3

v2.2

What's Changed

• Add NanoFlow code link by @DefTruth in #51
• 🔥[ACTIVATION SPARSITY] TRAINING-FREE ACTIVATION SPARSITY IN LARGE LANGUAGE MODELS by @DefTruth in #52
• 🔥[Decentralized LLM] Decentralized LLM Inference over Edge Networks with Energy Harvesting by @DefTruth in #53
• 🔥[SJF Scheduling] Efficient LLM Scheduling by Learning to Rank by @DefTruth in #54
• 🔥[Speculative Decoding] Boosting Lossless Speculative Decoding via Feature Sampling and Partial Alignment Distillation by @DefTruth in #55
• 🔥🔥[Prompt Compression] Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference by @DefTruth in #56
• 🔥🔥[Context Distillation] Efficient LLM Context Distillation by @DefTruth in #57
• Bump up to v2.2 by @DefTruth in #58

Full Changelog: v2.1...v2.2

v2.1

What's Changed

• Update README.md by @DefTruth in #40
• 🔥[Speculative Decoding] Parallel Speculative Decoding with Adaptive Draft Length by @DefTruth in #41
• 🔥[FocusLLM] FocusLLM: Scaling LLM’s Context by Parallel Decoding by @DefTruth in #42
• 🔥[NanoFlow] NanoFlow: Towards Optimal Large Language Model Serving Throughput by @DefTruth in #43
• 🔥[MagicDec] MagicDec: Breaking the Latency-Throughput Tradeoff for Long Context Generation with Speculative Decoding by @DefTruth in #44
• Add ABQ-LLM code link by @DefTruth in #46
• 🔥🔥[MARLIN] MARLIN: Mixed-Precision Auto-Regressive Parallel Inference on Large Language Models by @DefTruth in #47
• 🔥[1-bit LLMs] Matmul or No Matmal in the Era of 1-bit LLMs by @DefTruth in #48
• 🔥🔥[FLA] FLA: A Triton-Based Library for Hardware-Efficient Implementa… by @DefTruth in #49
• Bump up to v2.1 by @DefTruth in #50

Full Changelog: v2.0...v2.1