torch.fx.proxy.TraceError:symbolically traced variables cannot be used as inputs to control flow

1

Today, I use get_graph_node_names to get the trian nodes and test nodes.

such as : train_nodes, test_nodes = get_graph_node_names(module)

But while I’m running it

There a error about this :

torch.fx.proxy.TraceError:symbolically traced variables cannot be used as inputs to control flow.

I know. This problem with the model，but the model can be normally forward

I can’t find other problem from this model.

please, help me, how to deal it. :handshake:

the code follow as :

from torch.fx import Tracer
import torch.nn as nn
import torch

lstm = nn.LSTM(300, 100, 1)
x = torch.randn(7, 64, 300)
h = torch.randn(1, 64, 100)
c = torch.randn(1, 64, 100)
lstm(x, (h, c))

nodes = Tracer().trace(lstm).nodes
1 Like