new_impl/cv/dnns/vit_old/trans_weight.txt

cls_token	torch.Size([1, 1, 768])
mlp_head.0.bias	torch.Size([768])
mlp_head.0.weight	torch.Size([768])
transformer.layers.0.0.fn.norm.bias	torch.Size([768])
transformer.layers.0.0.fn.norm.weight	torch.Size([768])
transformer.layers.0.1.fn.norm.bias	torch.Size([768])
transformer.layers.0.1.fn.norm.weight	torch.Size([768])
transformer.layers.0.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.0.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.0.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.0.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.0.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.0.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.0.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.0.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.1.0.fn.norm.bias	torch.Size([768])
transformer.layers.1.0.fn.norm.weight	torch.Size([768])
transformer.layers.1.1.fn.norm.bias	torch.Size([768])
transformer.layers.1.1.fn.norm.weight	torch.Size([768])
transformer.layers.1.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.1.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.1.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.1.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.1.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.1.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.1.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.1.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.10.0.fn.norm.bias	torch.Size([768])
transformer.layers.10.0.fn.norm.weight	torch.Size([768])
transformer.layers.10.1.fn.norm.bias	torch.Size([768])
transformer.layers.10.1.fn.norm.weight	torch.Size([768])
transformer.layers.10.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.10.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.10.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.10.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.10.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.10.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.10.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.10.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.11.0.fn.norm.bias	torch.Size([768])
transformer.layers.11.0.fn.norm.weight	torch.Size([768])
transformer.layers.11.1.fn.norm.bias	torch.Size([768])
transformer.layers.11.1.fn.norm.weight	torch.Size([768])
transformer.layers.11.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.11.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.11.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.11.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.11.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.11.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.11.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.11.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.2.0.fn.norm.bias	torch.Size([768])
transformer.layers.2.0.fn.norm.weight	torch.Size([768])
transformer.layers.2.1.fn.norm.bias	torch.Size([768])
transformer.layers.2.1.fn.norm.weight	torch.Size([768])
transformer.layers.2.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.2.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.2.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.2.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.2.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.2.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.2.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.2.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.3.0.fn.norm.bias	torch.Size([768])
transformer.layers.3.0.fn.norm.weight	torch.Size([768])
transformer.layers.3.1.fn.norm.bias	torch.Size([768])
transformer.layers.3.1.fn.norm.weight	torch.Size([768])
transformer.layers.3.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.3.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.3.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.3.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.3.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.3.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.3.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.3.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.4.0.fn.norm.bias	torch.Size([768])
transformer.layers.4.0.fn.norm.weight	torch.Size([768])
transformer.layers.4.1.fn.norm.bias	torch.Size([768])
transformer.layers.4.1.fn.norm.weight	torch.Size([768])
transformer.layers.4.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.4.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.4.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.4.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.4.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.4.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.4.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.4.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.5.0.fn.norm.bias	torch.Size([768])
transformer.layers.5.0.fn.norm.weight	torch.Size([768])
transformer.layers.5.1.fn.norm.bias	torch.Size([768])
transformer.layers.5.1.fn.norm.weight	torch.Size([768])
transformer.layers.5.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.5.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.5.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.5.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.5.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.5.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.5.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.5.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.6.0.fn.norm.bias	torch.Size([768])
transformer.layers.6.0.fn.norm.weight	torch.Size([768])
transformer.layers.6.1.fn.norm.bias	torch.Size([768])
transformer.layers.6.1.fn.norm.weight	torch.Size([768])
transformer.layers.6.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.6.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.6.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.6.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.6.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.6.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.6.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.6.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.7.0.fn.norm.bias	torch.Size([768])
transformer.layers.7.0.fn.norm.weight	torch.Size([768])
transformer.layers.7.1.fn.norm.bias	torch.Size([768])
transformer.layers.7.1.fn.norm.weight	torch.Size([768])
transformer.layers.7.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.7.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.7.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.7.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.7.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.7.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.7.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.7.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.8.0.fn.norm.bias	torch.Size([768])
transformer.layers.8.0.fn.norm.weight	torch.Size([768])
transformer.layers.8.1.fn.norm.bias	torch.Size([768])
transformer.layers.8.1.fn.norm.weight	torch.Size([768])
transformer.layers.8.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.8.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.8.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.8.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.8.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.8.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.8.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.8.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
transformer.layers.9.0.fn.norm.bias	torch.Size([768])
transformer.layers.9.0.fn.norm.weight	torch.Size([768])
transformer.layers.9.1.fn.norm.bias	torch.Size([768])
transformer.layers.9.1.fn.norm.weight	torch.Size([768])
transformer.layers.9.1.fn.fn.net.0.bias	torch.Size([3072])
transformer.layers.9.1.fn.fn.net.0.weight	torch.Size([3072, 768])
transformer.layers.9.1.fn.fn.net.3.bias	torch.Size([768])
transformer.layers.9.1.fn.fn.net.3.weight	torch.Size([768, 3072])
transformer.layers.9.0.fn.fn.to_out.0.bias	torch.Size([768])
transformer.layers.9.0.fn.fn.to_out.0.weight	torch.Size([768, 768])
transformer.layers.9.0.fn.fn.to_qkv.bias	torch.Size([2304])
transformer.layers.9.0.fn.fn.to_qkv.weight	torch.Size([2304, 768])
pos_embedding	torch.Size([1, 197, 768])
to_patch_embedding.0.bias	torch.Size([768])
to_patch_embedding.0.weight	torch.Size([768, 3, 16, 16])
mlp_head.1.bias	torch.Size([21843])
mlp_head.1.weight	torch.Size([21843, 768])
pre_logits.fc.bias	torch.Size([768])
pre_logits.fc.weight	torch.Size([768, 768])