zengxiao-he/tessera

tessera

✅ Recently maintained

From teacher to tiles — a from-scratch LLM distillation & serving engine: custom Triton/CUDA kernels, FSDP distillation, paged-KV continuous batching, speculative decoding, a Rust gateway, a JAX oracle, and interpretability tooling.

558Stars
8Forks
07-day stars

Python · License unverified · Last push 2026-10-09

Official repository & installation ↗
cudaflash-attentionfsdpinference-enginejaxknowledge-distillationkv-cachellmmechanistic-interpretabilityml-systems

GitValue · Value evidence

Unscored · Evidence coverage 0%

Provisional rubric; documentation and runtime testing are separate. Missing evidence does not imply low value.

Star history

2026-06-012026-11-03
View historical data
DateStars
2026-10-05557
2026-10-06557
2026-10-07557
2026-10-08557
2026-10-09557
2026-10-10557
2026-10-11557
2026-10-12557
2026-10-13557
2026-10-14557
2026-10-15557
2026-10-16557
2026-10-17557
2026-10-18557
2026-10-19557
2026-10-20557
2026-10-21557
2026-10-22557
2026-10-23557
2026-10-24558
2026-10-25558
2026-10-26558
2026-10-27558
2026-10-28558
2026-10-29558
2026-10-30558
2026-10-31558
2026-11-01558
2026-11-02558
2026-11-03558

Get the project

git clone https://github.com/zengxiao-he/tessera.git

Cloning downloads the source. For dependencies and installation, follow the project README.

Related projects & alternative candidates

Related domains do not imply feature parity. Review official documentation and licenses.

vllm

High-throughput, memory-efficient LLM inference and serving engine that lets developers deploy models faster without building serving infrastructure.

★ 96,180sglang

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

★ 38,386