Open ArulselvanMadhavan opened 2 years ago
Describe the bug Pytorch 1.12 - torchvision - NonDynamicallyQuantizableLinear Not producing input/output shapes for this layer. I have pasted sample code and sample output below
class MultiheadAttentionWrapper(torch.nn.Module): def __init__(self, layer): super().__init__() self.layer = layer def forward(self, inputs): output = self.layer(inputs, inputs, inputs) return output[0] def __getitem__(self, index): return self.layer model = MultiheadAttentionWrapper(torch.nn.MultiheadAttention(768,12,batch_first=True)) col_names = ( "input_size", "output_size", "num_params", ) orig_results = torchinfo.summary(model, input_data=[input_tensors], verbose=0, device=device, col_names = col_names) print(orig_results)
Expected behavior A clear and concise description of what you expected to happen.
[============================================================================================================================= Layer (type:depth-idx) Input Shape Output Shape Param # ============================================================================================================================= MultiheadAttentionWrapper [2, 197, 768] [2, 197, 768] -- ├─MultiheadAttention: 1-1 [2, 197, 768] [2, 197, 768] 1,771,776 │ └─NonDynamicallyQuantizableLinear: 2-1 -- -- 590,592 ============================================================================================================================= Total params: 2,362,368 Trainable params: 2,362,368 Non-trainable params: 0 Total mult-adds (M): 0.00 ============================================================================================================================= Input size (MB): 1.21 Forward/backward pass size (MB): 0.00 Params size (MB): 2.36 Estimated Total Size (MB): 3.57 =============================================================================================================================]
Desktop (please complete the following information):
Additional context Add any other context about the problem here.
What are the expected input/output values?
Describe the bug Pytorch 1.12 - torchvision - NonDynamicallyQuantizableLinear Not producing input/output shapes for this layer. I have pasted sample code and sample output below
Expected behavior A clear and concise description of what you expected to happen.
Desktop (please complete the following information):
Additional context Add any other context about the problem here.