SemlaFlow is a small-molecule generation model from AstraZeneca using equivariant attention-based...
91
SemlaFlow is a small-molecule generation model from AstraZeneca using equivariant attention-based message passing (Semla architecture) with flow matching and optimal transport. The model generates 3D molecular structures; this AIM returns the canonical SMILES of each generated molecule together with its 3D structure as a mol block with 3D coordinates. Trained on GEOM-Drugs and QM9 datasets. The upstream code is MIT licensed; the checkpoint and data are fetched by the user from the upstream links, which state no separate licence.
| Vendor | AMD |
| License | MIT |
| Source | https://github.com/AMD-SiloAI/aim-build |
| Revision | 1879365e09ea4482ceb923876800fe7983d9f077 |
| Built | 2026-11-11T18:40:18Z |
Content type
Image
Digest
sha256:6dfefe959…
Size
10.3 GB
Last updated
3 days ago
docker pull amdenterpriseai/aim-instinct-model-rssrwn-semlaflow:2026.24.0-preview