ScalingIntelligence / hydragenLinks
Hydragen: High-Throughput LLM Inference with Shared Prefixes
☆43Updated last year
Alternatives and similar repositories for hydragen
Users that are interested in hydragen are comparing it to the libraries listed below
Sorting:
- [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding☆130Updated 10 months ago
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆169Updated last year
- TritonBench: Benchmarking Large Language Model Capabilities for Generating Triton Operators☆87Updated 4 months ago
- Boosting 4-bit inference kernels with 2:4 Sparsity☆84Updated last year
- PyTorch bindings for CUTLASS grouped GEMM.☆125Updated 5 months ago
- ☆102Updated 5 months ago
- ☆71Updated 7 months ago
- Framework to reduce autotune overhead to zero for well known deployments.☆84Updated last month
- 16-fold memory access reduction with nearly no loss☆104Updated 7 months ago
- ☆65Updated 6 months ago
- ☆145Updated 8 months ago
- [ICLR'25] Fast Inference of MoE Models with CPU-GPU Orchestration☆238Updated 11 months ago
- Triton-based Symmetric Memory operators and examples☆48Updated 2 weeks ago
- ☆50Updated 5 months ago
- ☆79Updated 6 months ago
- How to ensure correctness and ship LLM generated kernels in PyTorch☆107Updated last week
- Odysseus: Playground of LLM Sequence Parallelism☆78Updated last year
- [NeurIPS 2024] Efficient LLM Scheduling by Learning to Rank☆61Updated 11 months ago
- ☆112Updated last year
- [ICLR 2025] TidalDecode: A Fast and Accurate LLM Decoding with Position Persistent Sparse Attention☆48Updated 2 months ago
- Transformers components but in Triton☆34Updated 5 months ago
- [NeurIPS 2025] Scaling Speculative Decoding with Lookahead Reasoning☆51Updated this week
- Code for paper: [ICLR2025 Oral] FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference☆148Updated 2 weeks ago
- [NeurIPS 2024] The official implementation of "Kangaroo: Lossless Self-Speculative Decoding for Accelerating LLMs via Double Early Exitin…☆60Updated last year
- ☆130Updated 5 months ago
- 🔥 LLM-powered GPU kernel synthesis: Train models to convert PyTorch ops into optimized Triton kernels via SFT+RL. Multi-turn compilation…☆89Updated 3 weeks ago
- QuTLASS: CUTLASS-Powered Quantized BLAS for Deep Learning☆120Updated last week
- GPU operators for sparse tensor operations☆35Updated last year
- Estimate MFU for DeepSeekV3☆26Updated 9 months ago
- ☆58Updated this week