deepseek-ai/DeepSelect

Trend Velocity Score: 0.00
Opportunity Score: 50.00
Confidence Score: 20.00
Language: <a class="text-blue-600 dark:text-blue-400 hover:underline" href="/languages/Cuda">Cuda
License: MIT
First Seen: 4 weeks ago

Skip(New discovery — enrichment pending)

DeepSelect: TopK kernels for DeepSeek Sparse Attention (DSA) and Samplers

Why this repo matters

Implements a high-performance TopK kernel for DeepSeek models, offering significant speedups.
Focuses on specific scenarios for `bfloat16` and `float32` with optimizations for batch and vocab sizes.
Provides a drop-in replacement for `torch.topk` with potential performance gains.

The agent that grows with you

DeepSeek Harness: Everything is a Plugin.

Use Garry Tan's exact Claude Code setup: 23 opinionated tools that serve as CEO, Designer, Eng Manager, Release Manager, Doc Engineer, and QA

An AI skill that provides design intelligence for building professional UI/UX across multiple platforms.