lmxmercy / MGCT

Mutual-Guided Cross-Modality Transformer
4 stars 1 forks source link

Inquiry about S1 and S2 parameter issues #4

Open Ge-YL opened 1 month ago

Ge-YL commented 1 month ago

`

Stage 1: WSI features and Genomic features fusion

    for stage1_layer in self.stage1_layers: # 1
        stage1_output, omic_wsi_mgca = stage1_layer(wsi_feats_embed, omic_feats_embed, attn_mask=None)

Stage 2: stage1 fusion embedding and WSI feature fusion again

    for stage2_layer in self.stage2_layers: # 2
        stage2_output, wsi_omic_mgca = stage2_layer(stage1_output.unsqueeze(0).unsqueeze(0), wsi_feats_embed, attn_mask=None)

` According to your framework, the second for loop has two layers, S2=2. But from the code perspective, it seems that we can only use the result of the second iteration, so what is the significance of doing the first loop of S2?

Ge-YL commented 1 month ago

@lmxmercy