From-scratch reimplementation of DeepSeek's Native Sparse Attention (arXiv:2502.11089) in Triton + CUDA Hopper WGMMA. 7.4x faster than FlashAttention-3 at 64k context. Five-model training fleet, perplexity sweep, LongBench v2, MoBA comparison.
GitHub repository with 6 stars and 0 forks.
Language: Python
Topics: attention-mechanism, cuda, deepseek, flash-attention, gpu-kernels, hopper, llm, long-context, native-sparse-attention, nsa
Trending score 0.00, activity score 0.00, stars gained +0, forks gained +0.
2026-06-05: 6 stars and 0 forks.
A concise but complete full-attention transformer with a set of promising experimental features from various papers
GitHub repository with 5,887 stars and 510 forks.
Trending score: 0.68; stars gained: +2; forks gained: +0.
Language: Python
Topics: artificial-intelligence, deep-learning, attention-mechanism, transformers
📝 Streamline text processing in Arabic and English with ChunkWise, a library offering 31 chunking strategies for NLP and RAG systems.
GitHub repository with 5 stars and 0 forks.
Trending score: 0.05; stars gained: +0; forks gained: +0.
Language: Python
Topics: aggregation, attention-mechanism, chunk, chunkwise-processing, csv, data-analysis
From-scratch reimplementation of DeepSeek's Native Sparse Attention (arXiv:2502.11089) in Triton + CUDA Hopper WGMMA. 7.4x faster than FlashAttention-3 at 64k context. Five-model training fleet, perplexity sweep, LongBench v2, MoBA comparison.
GitHub repository with 6 stars and 0 forks.
Trending score: 0.00; stars gained: +0; forks gained: +0.
Language: Python
Topics: attention-mechanism, cuda, deepseek, flash-attention, gpu-kernels, hopper
The agent that grows with you
GitHub repository with 182,513 stars and 31,295 forks.
Trending score: 5.95; stars gained: +1,867; forks gained: +361.
Language: Python
Topics: ai, ai-agent, ai-agents, anthropic, chatgpt, claude
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 60-95% fewer tokens, same answers. Library, proxy, MCP server.
GitHub repository with 14,053 stars and 885 forks.
Trending score: 5.69; stars gained: +2,829; forks gained: +175.
Language: Python
Topics: agent, ai, anthropic, compression, context-engineering, context-window
Academic Research Skills for Claude Code: research → write → review → revise → finalize
GitHub repository with 27,616 stars and 2,272 forks.
Trending score: 5.52; stars gained: +1,079; forks gained: +89.
Language: Python
Topics: academic-pipeline, academic-writing, ai-research, claude, claude-code, literature-review
Learn it. Build it. Ship it for others.
GitHub repository with 28,711 stars and 4,695 forks.
Trending score: 5.32; stars gained: +1,261; forks gained: +238.
Language: Python
Topics: agents, ai, ai-agents, ai-engineering, computer-vision, course
An opinionated list of Python frameworks, libraries, tools, and resources
GitHub repository with 301,435 stars and 28,046 forks.
Trending score: 4.60; stars gained: +518; forks gained: +24.
Language: Python
Topics: awesome, collections, python, python-frameworks, python-libraries, python-tools
Use claude-code for free in the terminal, VSCode extension or discord like OpenClaw (voice supported)
GitHub repository with 32,540 stars and 4,942 forks.
Trending score: 4.56; stars gained: +467; forks gained: +82.
Language: Python
A concise but complete full-attention transformer with a set of promising experimental features from various papers
GitHub repository with 5,887 stars and 510 forks.
Trending score: 0.68; stars gained: +2; forks gained: +0.
Language: Python
Topics: artificial-intelligence, deep-learning, attention-mechanism, transformers
邱锡鹏《神经网络与深度学习》（蒲公英书）理论书 v2 与通识版
GitHub repository with 18,809 stars and 3,667 forks.
Trending score: 0.29; stars gained: +0; forks gained: +1.
Topics: attention-mechanism, chinese, deep-learning, machine-learning, neural-networks, textbook
📝 Streamline text processing in Arabic and English with ChunkWise, a library offering 31 chunking strategies for NLP and RAG systems.
GitHub repository with 5 stars and 0 forks.
Trending score: 0.05; stars gained: +0; forks gained: +0.
Language: Python
Topics: aggregation, attention-mechanism, chunk, chunkwise-processing, csv, data-analysis
From-scratch reimplementation of DeepSeek's Native Sparse Attention (arXiv:2502.11089) in Triton + CUDA Hopper WGMMA. 7.4x faster than FlashAttention-3 at 64k context. Five-model training fleet, perplexity sweep, LongBench v2, MoBA comparison.
GitHub repository with 6 stars and 0 forks.
Trending score: 0.00; stars gained: +0; forks gained: +0.
Language: Python
Topics: attention-mechanism, cuda, deepseek, flash-attention, gpu-kernels, hopper