Search Results for "throughput"

NVIDIA® TensorRT™ is an SDK for high-performance deep learning inference. It includes a deep learning inference optimizer and runtime that delivers low latency and high throughput for deep learning inference applications. TensorRT-based applications perform up to 40X faster than CPU-only platforms during inference. With TensorRT, you can optimize neural network models trained in all major frameworks, calibrate for lower precision with high accuracy, and deploy to hyperscale data centers, embedded, or automotive product platforms. ...

FlashMLA is a high-performance decoding kernel library designed especially for Multi-Head Latent Attention (MLA) workloads, targeting NVIDIA Hopper GPU architectures. It provides optimized kernels for MLA decoding, including support for variable-length sequences, helping reduce latency and increase throughput in model inference systems using that attention style. The library supports both BF16 and FP16 data types, and includes a paged KV cache implementation with a block size of 64 to efficiently manage memory during decoding. On very compute-bound settings, it can reach up to ~660 TFLOPS on H800 SXM5 hardware, while in memory-bound configurations it can push memory throughput to ~3000 GB/s. ...

...Utterly ignores even the most powerful GPU in your system and uses only one CPU core per prediction. Quite fast on one CPU core, and you can run multiple predictions in parallel, thus utilizing as many CPUs as you like to improve the overall prediction throughput of your application/pipeline.

...Transactional semantics allow atomic updates and consistent snapshots, which is essential for large, shared datasets used by ML and scientific workflows. The library is engineered for scalability—background caching, chunk sharding, and retryable operations keep throughput high even over unreliable networks. With language bindings, it fits into Python-heavy analysis pipelines while retaining a fast C++ core.

...It’s designed to defeat hash-flooding attacks by mixing input with wide SIMD operations and a branch-free inner loop, so adversaries can’t cheaply craft many colliding keys. The implementation targets multiple CPU families with vectorized code paths while keeping a portable fallback, yielding high throughput across platforms. It exposes simple one-shot and streaming APIs, so you can hash short keys or long byte streams with the same function. Typical uses include protecting hash tables that store untrusted keys and authenticating short-lived messages or records in storage systems. Although not a replacement for collision-resistant digests like SHA-2/3, it strikes a pragmatic balance of speed, simplicity, and resistance to common abuse patterns seen in production backends.