OpenFold is trainable in full precision or `bfloat16` with or without DeepSpeed,
OpenFold is trainable in full precision or `bfloat16` with or without DeepSpeed,
and we've trained it from scratch, matching the performance of the original.
and we've trained it from scratch, matching the performance of the original.
We've publicly released model weights and our training data &mdash some 400,000
We've publicly released model weights and our training data — some 400,000
MSAs &mdash under a permissive license. Model weights are available from
MSAs — under a permissive license. Model weights are available from
scripts in this repository while the MSAs are hosted by the
scripts in this repository while the MSAs are hosted by the
[Registry of Open Data on AWS (RODA)](registry.opendata.aws/openfold).
[Registry of Open Data on AWS (RODA)](registry.opendata.aws/openfold).
Try out running inference for yourself with our [Colab notebook](https://colab.research.google.com/github/aqlaboratory/openfold/blob/main/notebooks/OpenFold.ipynb).
Try out running inference for yourself with our [Colab notebook](https://colab.research.google.com/github/aqlaboratory/openfold/blob/main/notebooks/OpenFold.ipynb).
