I've got the following exception: ValueError: Expected more than 1 spatial element when training, got input size torch.Size([1, 512, 1]) on the model forward pass (z, log_s_list, gate_pred, attn, attn_logprob, mean, log_var, prob = model(mel, spk_ids, txt, in_lens, out_lens, attn_prior))
I've discovered that this exception occurs if the whole text (txt) is a single word (e.g. "what"). I found some mentions of this problem in this repo's issues but no solution was provided.
The full traceback:
File "/slot/sandbox/nv_tmpfs/d/in/script/0_script_unpacked/s2t_probs/flowtron_score_text.py", line 147, in <module>
attn_logprob, mean, log_var, prob) = model(
File "/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/slot/sandbox/nv_tmpfs/d/in/script/0_script_unpacked/s2t_probs/flowtron/flowtron.py", line 818, in forward
text = self.encoder(text, in_lens)
File "/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/slot/sandbox/nv_tmpfs/d/in/script/0_script_unpacked/s2t_probs/flowtron/flowtron.py", line 436, in forward
F.relu(conv(curr_x)),
File "/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/anaconda3/lib/python3.8/site-packages/torch/nn/modules/container.py", line 141, in forward
input = module(input)
File "/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/anaconda3/lib/python3.8/site-packages/torch/nn/modules/instancenorm.py", line 57, in forward
return F.instance_norm(
File "/anaconda3/lib/python3.8/site-packages/torch/nn/functional.py", line 2326, in instance_norm
_verify_spatial_size(input.size())
File "/anaconda3/lib/python3.8/site-packages/torch/nn/functional.py", line 2293, in _verify_spatial_size
raise ValueError("Expected more than 1 spatial element when training, got input size {}".format(size))
ValueError: Expected more than 1 spatial element when training, got input size torch.Size([1, 512, 1])
I understand why processing a sequence of a length 1 through an instance norm causes an exception. However, is there a way to make it work? I mean, maybe I could add some special token to the string or something like that.
Hello!
I've got the following exception:
ValueError: Expected more than 1 spatial element when training, got input size torch.Size([1, 512, 1])
on the model forward pass (z, log_s_list, gate_pred, attn, attn_logprob, mean, log_var, prob = model(mel, spk_ids, txt, in_lens, out_lens, attn_prior)
)I've discovered that this exception occurs if the whole text (
txt
) is a single word (e.g."what"
). I found some mentions of this problem in this repo's issues but no solution was provided.The full traceback:
I understand why processing a sequence of a length 1 through an instance norm causes an exception. However, is there a way to make it work? I mean, maybe I could add some special token to the string or something like that.