Pinned Loading

1. General-Instinct/InstinctFlash General-Instinct/InstinctFlash Public

   High-Performance Serving Runtime for Robotics Models

   Python 199 6

2. sparse-attention-triton-bench sparse-attention-triton-bench Public

   Content-adaptive sparse attention in Triton. Runtime block selection, online softmax. Benchmarked vs FA2 and FlexAttention at 32k context.

   Python 1

3. multi-agent-memory-eval multi-agent-memory-eval Public

   temporal knowledge-graph memory vs. RAG on a multi-agent research system

   Python 1

4. mini-vllm mini-vllm Public

   From-scratch LLM inference engine with ablation testing

   Python

5. bpe-tokenizer bpe-tokenizer Public

   Byte Pair Encoding tokenization from scratch in Python

   Jupyter Notebook

6. gpt_from_scratch gpt_from_scratch Public

   GPT-2 style transformer built from scratch in PyTorch

   Jupyter Notebook 1