globus-labs / mof-generation-at-scale / 8522963786

46%

Build Type

push

github

Committed by web-flow

Commit Message

Add retraining DiffLinker to the workflow (#81)

* Move difflinker training item generation to model

* First pass at using examples in training

1. Create a JSON.gz file in before calling train script
2. Set the dataset override flag
3. Fix a few iterations

* Use a model from checkpoint

* Start a component test

* Update to use Xiaoli's latest model

* First attempt at XPU support

* Implement abstract methods

* Optimize the optimizer

* Add a hook for moving optimizer and model to XPU

* Use the GPU, but only one worker

Ensures that we don't get two pytorch tasks on the GPU
at the same time

* Add retraining to the workflow

* Add test file with 1024 example MOFs

* Reduce training set size

Run Details

97 of 114 new or added lines in 7 files covered. (85.09%)

229 existing lines in 5 files now uncovered.

4229 of 10061 relevant lines covered (42.03%)

0.42 hits per line