moritztngmoritztngCommunity·1 Agent Skills Included

tt-bio

Protein structure prediction and design on Tenstorrent hardware

Runs protein structure prediction, binder design, and binding affinity models like Boltz-2, ESMFold2, Protenix, and OpenFold3 on Tenstorrent hardware. Eliminates dependence on expensive NVIDIA GPU clusters by supporting single-card, multi-card, and multi-machine configurations. Validates every model against its official reference implementation and publishes throughput and cost benchmarks for each.
npx skills add moritztng/tt-bio --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install tt-bio?

Run `npx skills add moritztng/tt-bio --all -g -y` in your terminal to install all skills in this suite globally.

What does tt-bio do?

It runs protein structure prediction, binder design, and binding affinity models such as Boltz-2, ESMFold2, Protenix, and OpenFold3 on Tenstorrent Blackhole and Wormhole hardware.

Can tt-bio run without Tenstorrent hardware?

Yes. The Boltz-2 CPU and GPU path and the command-line interface work on a standard host, while the other models require a Tenstorrent card.

How accurate are tt-bio predictions?

Every model is validated against its official reference implementation on the same input and reproduces results within that reference's own run-to-run noise.

Does tt-bio support multi-card setups?

Yes. It supports single-card and multi-card configurations like a 4-card QuietBox or 32-card Galaxy server, and multiple machines can be combined into one prediction run.

Related Repositories in Education & Research

View All in Education & Research