A question

[ccyniubi]

I want to know that this in readme.md:

For example, to run OGM on CREMA-D dataset:

python -m balancemm \
    --trainer OGM \
    --dataset CREMAD \
    --model BaseClassifier \
    --alpha 0.5 \
    --device 0

The CREMAD dataset is a bimodal dataset, but BaseClassifier is trimodal like:

model:
  BaseClassifier:
    encoders: {audio: Transformer_, visual: Transformer_, text:  Transformer_}
    fusion: "concat"
    modality_size: 1024

Is there any problem here?
Also, the paper mentions that the FOOD101 dataset uses a model shared by resnet and transformer, but this does not seem to be in model_config.yaml.
Looking forward to your reply.