Tutorials #8
Tutorials #8
Conversation
Co-authored-by: Marie Mueller <vc340@users.noreply.github.com>
ToKa96
left a comment
ToKa96
left a comment
There was a problem hiding this comment.
Looks great! Runs through with a fresh environment. A few warnings/errors appear during visualization on my machine but they do not seem to imper the visualization itself
| " # we need one simple override here but otherwise we just use the default setting (see tutorial 4 for more information)\n", | ||
| " \"data.dataset_name=QM9_perturbed_fock\", # this will no longer be necessary once the \"fixed\" is removed from the dataset_name\n", |
There was a problem hiding this comment.
I think we have already removed the fixed and do not need it here
| " # we need one simple override here but otherwise we just use the default setting (see tutorial 4 for more information)\n", | ||
| " \"data.dataset_name=QM9_perturbed_fock\", # this will no longer be necessary once the \"fixed\" is removed from the dataset_name\n", |
| " overrides=[\n", | ||
| " # we need one simple override here but otherwise we just use the default setting (see tutorial 4 for more information)\n", | ||
| " \"data.dataset_name=QM9_perturbed_fock\", # this will no longer be necessary once the \"fixed\" is removed from the dataset_name\n", |
| " # we need one simple override here but otherwise we just use the default setting (see tutorial 4 for more information)\n", | ||
| " \"data.dataset_name=QM9_perturbed_fock\", # this will no longer be necessary once the \"fixed\" is removed from the dataset_name\n", |
| " config_name=\"train.yaml\",\n", | ||
| " overrides=[\n", | ||
| " \"data/transforms=local_frames_global_natrep_add_lframes\",\n", | ||
| " \"data.dataset_name=QM9_perturbed_fock\", # this will no longer be necessary once the \"fixed\" is removed from the dataset_name\n", |
| "outputs": [], | ||
| "source": [ | ||
| "# change max number of interations:\n", | ||
| "density_optimizer.max_cycle = 10 # You migth want to change the number of iterations to speed things up, but this migth not guarantee convergence\n", |
There was a problem hiding this comment.
I would a) add a warning that with this number it will most likely not be converged, and b) add an estimate how many in our experience usually are sufficient. (Will always the same molecule be loaded here? If so I guess a good estimate would be around 500 cycles
| "# Note though that the displayed geometry is inferred from the SMILES string, so it does not exactly correspond to the geometry in the sample object.\n", | ||
| "# if you click on the link you will see 3D structure of the molecule in the browser:\n", |
There was a problem hiding this comment.
If with the shipped state of the notebook the same molecule will always be loaded than the smiles and the xyz geometry will always differ in our example here, in this case I would actually tell the people to compare the two geometry and spot the difference to point that out
2 participants
What does this PR do?
This PR adds several notebook tutorials for different topics and functionality covered in the code base.
Before submitting
pytestcommand?pre-commit run -acommand?