I would like to add a module graph to tensorboard using
SummaryWriter.add_graph. My module’s
forward method requires some
*args and some
Walking down the stack, it looks like
torch.jit.get_trace_graph(model, args), i.e. not passing any kwargs, even though
torch.jit.get_trace_graph would support it.
Is there a way to pass kwargs? The only solution I see right now is
foward parameters and rearranging the arguments so that they are all converted to args, which I would really like to avoid.