Popular repositories Loading

1. nanochat-rs-ternary nanochat-rs-ternary Public

   Production-ready ternary quantized (1.58-bit) Rust code generation model with mHC-lite, MaxRL training, and comprehensive benchmarking

   Rust 3

2. sglang_attentio sglang_attentio Public

   We added top-k attention token visualization, exposing which tokens the model attends to most during decode, useful for interpretability research for moe hybrid mamba arch

3. sglang sglang Public

   Forked from sgl-project/sglang

   SGLang is a high-performance serving framework for large language models and multimodal models.

   Python

4. 30b-moe-benchmarks 30b-moe-benchmarks Public

   Comprehensive benchmarks for 30B MoE models on RTX 4090. Key finding: HumanEval ≠ SWE-bench - the 98% HumanEval model scores 0% on agentic tasks. Includes Nemotron, GLM-4.7-Flash, Qwen3-Coder compa…

   Python

5. selfware selfware Public

   Your personal AI workshop — software you own, software that lasts

   Rust