NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
https://nvidia.github.io/OpenSeq2Seq
Apache License 2.0
1.55k stars 369 forks source link

Could not test with infer for text2speech #527

Closed ghost closed 4 years ago

ghost commented 4 years ago

I've trained with text2speech/tacotron_float.py for my TTS model. It's done and I got checkpoint in log dir. But when I run mode=infer, it gives following error.

Processed 1/1 batches Processed 1/1 batches *** Not enough steps for benchmarking output_file is ignored for tts results are logged to the logdir ... ValueError: operands could not be broadcast together with shapes (401,350) (513,273)

This is my test.csv file and running command. csv file: UNUSED | UNUSED | This is an example sentence what I want to generate. comnand: python run.py --config_file=example_configs/text2speech/tacotron_float.py --mode=infer --infer_output_file=unused