TorchCraft
TorchCraft designs binder sequences by optimising a trainable amino-acid logit matrix through a frozen all-atom structure predictor and then discretising it into a sequence with its predicted complex.
Results are recorded on linked configurations and versions of this pipeline. Their settings and evaluations are kept separate below.
Results on linked records
This pipeline has no results of its own in this release. 6 linked records have results. They are not assigned to the underlying pipeline: each record keeps its own settings and evaluations, and its scores are on its own page.
- TorchCraft minibinder design, raw output Configuration · configuration of this pipeline · 4 evaluations · 4 results
- TorchCraft on A100 with cuequivariance kernels Configuration · configuration of this pipeline · 5 evaluations · 5 results
- TorchCraft on A100 without cuequivariance kernels Configuration · configuration of this pipeline · 5 evaluations · 5 results
- TorchCraft on Ascend 910B with fused attention Configuration · configuration of this pipeline · 5 evaluations · 5 results
- TorchCraft on Ascend 910B without fused attention Configuration · configuration of this pipeline · 5 evaluations · 5 results
- TorchCraft framework-conditioned VHH design, raw output Configuration · configuration of this pipeline · 4 evaluations · 4 results
Overview
An explanatory profile has not yet been reviewed for this record.
Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.
Model Capabilities & Hardware Requirements
Deployment facts recorded for TorchCraft. Fields the record does not carry are marked not recorded.
Evaluated Checkpoint: As published in arXiv:2609.19770v1; GPU tree at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd
Use this pipeline
How it works, versions and access
Additional methodological details, versions and access terms have not yet been extracted into a structured profile for this pipeline. Consult the linked sources and evaluation records below.
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
0 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|
No evidence rows match these filters. Choose another scope or clear the search.
Sources and history
Release 2026-11-21-c8bea25e5abb · Record review: source checked
4 source records and release history
- TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) · Original source · arXiv:2609.19770v1, submitted 17 September 2026
- TorchX repository README at commit ebf9dc6 · Original source · commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd, authored 2026-11-04T07:22:24Z
- TorchX torchcraft_gpu/README.md at commit ebf9dc6 · Original source · commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd
- TorchX project site (torchx-cpl.github.io) · Original source · Project site as served on 2026-11-15
Technical metadata and record identifiers
Technical metadata for this record is tracked under stable ID torchcraft-20261115-pipeline-torchcraft.
Source locator: Abstract; Sections 2.1 and 2.2; Appendix A.1.1 to A.1.4
Related records
- uses model: AlphaFold 3
- uses model: TorchFold
- subject: TorchCraft: Binding does not establish the predicted interface is correct
- subject: TorchCraft: The shared four-stage optimisation schedule
- subject: TorchCraft: Every reported TorchCraft experiment used pretrained AlphaFold 3 weights
- subject: TorchCraft: Experimental validation covers minibinders and VHHs only
- subject: TorchCraft: Implementation changes that make repeated predictor evaluation affordable
- subject: TorchCraft: Nothing in this batch was reproduced
- subject: TorchCraft: TorchCraft trains no model weights of its own
- subject: TorchCraft: The tested sequences were raw TorchCraft outputs
- configuration of: TorchCraft disulfide-linked cyclic-peptide design with ProteinMPNN polishing
- configuration of: TorchCraft disulfide-linked cyclic-peptide design, raw output
- configuration of: TorchCraft head-to-tail cyclic-peptide design with ProteinMPNN polishing
- configuration of: TorchCraft head-to-tail cyclic-peptide design, raw output
- configuration of: TorchCraft ligand-conditioned pocket design with LigandMPNN redesign
- configuration of: TorchCraft ligand-conditioned pocket design, raw output
- configuration of: TorchCraft minibinder design with ProteinMPNN polishing
- configuration of: TorchCraft minibinder design, raw output
- configuration of: TorchCraft on A100 with cuequivariance kernels
- configuration of: TorchCraft on A100 without cuequivariance kernels
- configuration of: TorchCraft on Ascend 910B with fused attention
- configuration of: TorchCraft on Ascend 910B without fused attention
- configuration of: TorchCraft VHH design with AbMPNN polishing
- configuration of: TorchCraft framework-conditioned VHH design, raw output
How to cite
TorchCraft. Rewire biological model benchmark database, release 2026-11-21-c8bea25e5abb. https://benchmarks.rewirebio.io/database/pipeline/torchcraft-20261115-pipeline-torchcraft/
Release 2026-11-21-c8bea25e5abb. Rankings and values change between releases, so keep the release id in the citation. Cite the original publication as well for any value you reuse: each result links its source.
The compilation and annotations made by Rewire in this release are offered under CC BY 4.0; values and text taken from cited sources stay under the terms of those sources and are not covered by this licence. Data licence