config.json

{
  "architectures": [
    "HierarchicalTwoStreamTransformer"
  ],
  "model_type": "custom_transformer",
  "cnn_backbone": "mobilenetv3_large_100",
  "visual_feature_dim": 256,
  "pose_feature_dim": 256,
  "fusion_type": "gated",
  "num_transformer_layers": 1,
  "num_attention_heads": 4,
  "embedding_dim": 256,
  "num_classes": 8,
  "image_size": [128, 128],
  "num_frames": 30
}