vllm-project / vllm
A high-throughput LLM inference and serving engine featuring PagedAttention, continuous batching, and an OpenAI-compatible API server.
vllm-project / vllm
FareedKhan-dev / kimi-k3-in-c
A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.
Dao-AILab / flash-attention
Fast and memory-efficient exact attention
mxsm / rocketmq-rust
π Apache RocketMQ built in Rust π¦ β faster, safer, and more memory-efficient. Powering high-performance messaging for the AI era, with AI Agent, MCP, and AI-SRE capabilities.
uhop / stream-json
A micro-library of stream components for building custom JSON and JSONC processing pipelines with a minimal memory footprint β parse, filter, and transform JSON far larger than available memory with a SAX-inspired token API, on Node.js or Web Streams.
matsumotory / ngx_mruby
ngx_mruby - A Fast and Memory-Efficient Web Server Extension Mechanism Using Scripting Language mruby for nginx
HKUSTDial / flash-sparse-attention
Trainable fast and memory-efficient sparse attention
FlashML-org / flashlib
Fast and memory-efficient classical machine learning operators
Saganaki22 / ComfyUI-sol-attn
NVIDIA Sol-Attn for ComfyUI / Triton kernel on SM89 - SM121, with zero-copy MiniMax H3 nodes: memory-efficient attention, scheduled tau with graph preview, and feed-forward chunking. Measured 1.14β1.44Γ vs SageAttention and β37% MLP peak VRAM on H3
yfyeung / PrunedCTC
Memory-efficient CTC loss with exact loss and first-order gradient equivalence under vocabulary reduction, and activation memory that no longer scales linearly with vocabulary size.
NVIDIA-BioNeMo / BioNeMo-Inference-Runtime
Easy, fast, and memory-efficient structure prediction inference
natera-open-source / vep-rs
A fast, memory-efficient variant effect predictor written in Rust: a from-scratch reimplementation of the Ensembl Variant Effect Predictor (VEP) that takes the same inputs and flags, emits VEP's consequence vocabulary in VEP's output formats, and annotates population-scale callsets two orders of magnitude faster than the Perl implementation.
Newaiguy / ComfyUI-h3_sage_amd
AMD gfx12 (RX 9070/9070 XT) Memory-Efficient Sage Attention for MiniMax H3. Matches KJNodes' Nε‘ patch with native HIP kernel.