Open GZ82 opened 10 months ago
Hi, tasks.MultipleBinaryClassification
is designed for protein function classification tasks. You can refer to the GearNet repo and this tutorial.
@Oxer11 Hi thanks very much for your prompt reply!
I want to predict toxicity of chemicals, say 1 toxic 0 non-toxic, i.e., a binary classification, and inputs are SMILES
does tasks.MultipleBinaryClassification
fit in this application?
or I just use tasks.PropertyPrediction
and set num_class=1
?
sorry I do not understand what does task=[_ for _ in range(len(dataset.tasks))]
this mean in
task = tasks.MultipleBinaryClassification(gearnet, graph_construction_model=graph_construction_model, num_mlp_layer=3,
task=[_ for _ in range(len(dataset.tasks))], criterion="bce", metric=["auprc@micro", "f1_max"])
in your tutorial
Hi Anyone can provide an example for this task: tasks.MultipleBinaryClassification the setting for task argument is very different to that of task. PropertyPrediction the document () mentioned it is optional, or a list of int, tried both optional or list of [1,0,...]
my dataset.tasks is ["one_or_zero"] when run:
get following errors: