esradonmez / ims-toucan Goto Github PK

IMS-Toucan is a toolkit for teaching, training and using state-of-the-art Speech Synthesis models, developed at the Institute for Natural Language Processing (IMS), University of Stuttgart, Germany. Everything is pure Python and PyTorch based to keep it as simple and beginner-friendly, yet powerful as possible.

The PyTorch Modules of Tacotron2 and FastSpeech2 are taken from ESPnet, the PyTorch Modules of HiFiGAN are taken from the ParallelWaveGAN repository which are also authored by the brilliant Tomoki Hayashi.

For a version of the toolkit that includes TransformerTTS instead of Tacotron2 and MelGAN instead of HiFiGAN, check out the TransformerTTS and MelGAN branch. They are separated to keep the code clean, simple and minimal.

• As shown in this paper vocoders can be used to perform super-resolution and spectrogram inversion simultaneously. We added this to our HiFi-GAN vocoder. It now takes 16kHz spectrograms as input, but produces 48kHz waveforms.
• We officially introduced IMS Toucan in our contribution to the Blizzard Challenge 2021. Check out the bottom of the readme for a bibtex entry.
• Plenty of new features for low-resource and multi-linguality coming soon

Here are two sentences produced by Tacotron 2 combined with HiFi-GAN, trained on Nancy Krebs using this toolkit.

Here is some speech produced by FastSpeech2 and MelGAN trained on LJSpeech using this toolkit.

And here is a sentence produced by TransformerTTS and MelGAN trained on Thorsten using this toolkit.

Here is some speech produced by a multi-speaker FastSpeech2 with MelGAN trained on LibriTTS using this toolkit. Fans of the videogame Portal may recognize who was used as the reference speaker for this utterance.

Interactive Demo of our entry to the Blizzard Challenge 2021. This is based on an older version of the toolkit though. It uses FastSpeech2 and MelGAN as vocoder and is trained on 5 hours of Spanish.

To install this toolkit, clone it onto the machine you want to use it on (should have at least one GPU if you intend to train models on that machine. For inference, you can get by without GPU). Navigate to the directory you have cloned. We are going to create and activate a conda virtual environment to install the basic requirements into. After creating the environment, the command you need to use to activate the virtual environment is displayed. The commands below show everything you need to do.

conda create --prefix ./toucan_conda_venv --no-default-packages python=3.8

pip install --no-cache-dir -r requirements.txt

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html

If you want to use multi-speaker synthesis, you will need a speaker embedding function. IMS Toucan uses three such functions as an ensemble.

• dvector
• Speechbrain's ECAPA-TDNN
• Speechbrain's X-Vector

In the current version of the toolkit no further action should be required. When you are using multispeaker for the first time, it requires an internet connection to download the pretrained models though.

And finally you need to have espeak-ng installed on your system, because it is used as backend for the phonemizer. If you replace the phonemizer, you don't need it. On most Linux environments it will be installed already, and if it is not, and you have the sufficient rights, you can install it by simply running

apt-get install espeak-ng

To create a new pipeline to train a HiFiGAN vocoder, you only need a set of audio files. To create a new pipeline for a Tacotron2 you need audio files and corresponding text labels. To create a new pipeline for a FastSpeech2, you need audio files, corresponding text labels, and an already trained Tacotron2 model to estimate the duration information that FastSpeech 2 needs as input. Let's go through them in order of increasing complexity.

In the directory called Utility there is a file called file_lists.py. In this file you should write a function that returns a list of all the absolute paths to each of the audio files in your dataset as strings.

Then go to the directory TrainingInterfaces/TrainingPipelines. In there, make a copy of any existing pipeline that has HiFiGAN in its name. We will use this as reference and only make the necessary changes to use the new dataset. Import the function you have just written as get_file_list. Now look out for a variable called model_save_dir. This is the default directory that checkpoints will be saved into, unless you specify another one when calling the training script. Change it to whatever you like.

Now you need to add your newly created pipeline to the pipeline dictionary in the file run_training_pipeline.py in the top level of the toolkit. In this file, import the run function from the pipeline you just created and give it a speaking name. Now in the pipeline_dict, add your imported function as value and use as key a shorthand that makes sense. And just like that you're done.

In the directory called Utility there is a file called path_to_transcript_dicts.py. In this file you should write a function that returns a dictionary that has all the absolute paths to each of the audio files in your dataset as strings as the keys and the textual transcriptions of the corresponding audios as the values.

Then go to the directory TrainingInterfaces/TrainingPipelines. In there, make a copy of any existing pipeline that has Tacotron2 in its name. If your dataset is single-speaker, choose any that is not LibriTTS. If your dataset is multi-speaker, choose the one for LibriTTS as your template. We will use this copy as reference and only make the necessary changes to use the new dataset. Import the function you have just written as build_path_to_transcript_dict. Since the data will be processed a considerable amount, a cache will be built and saved as file for quick and easy restarts. So find the variable cache_dir and adapt it to your needs. The same goes for the variable save_dir, which is where the checkpoints will be saved to. This is a default value, you can overwrite it when calling the pipeline later using a command line argument, in case you want to fine-tune from a checkpoint and thus save into a different directory.

Since we are using text here, we have to make sure that the text processing is adequate for the language. So check in Preprocessing/TextFrontend whether the TextFrontend already has a language ID (e.g. 'en' and 'de') for the language of your dataset. If not, you'll have to implement handling for that, but it should be pretty simple by just doing it analogous to what is there already. Now back in the pipeline, change the lang argument in the creation of the dataset and in the call to the train loop function to the language ID that matches your data.

Now navigate to the implementation of the train_loop that is called in the pipeline. In this file, find the function called plot_attention. This function will produce attention plots during training, which is the most important way to monitor the progress of the training. In there, you may need to add an example sentence for the language of the data you are using. It should all be pretty clear from looking at it.

Once this is done, we are almost done, now we just need to make it available to the run_training_pipeline.py file in the top level. In said file, import the run function from the pipeline you just created and give it a speaking name. Now in the pipeline_dict, add your imported function as value and use as key a shorthand that makes sense. And that's it.

Most of this is exactly analogous to building a Tacotron2 pipeline. So to keep this brief, this section will only mention the additional things you have to do.

In your new pipeline file, look out for the line in which the acoustic_model is loaded. Change the path to the checkpoint of a Tacotron2 model that you trained on the same dataset previously. This is used to estimate phoneme-durations based on knowledge-distillation.

Everything else is exactly like creating a Tacotron2 pipeline, except that in the training_loop, instead of attentions plots, spectrograms are plotted to visualize training progress. So there you may need to add a sentence if you are using a new language in the function called plot_progress_spec.

Once you have a pipeline built, training is super easy. Just activate your virtual environment and run the command below. You might want to use something like nohup to keep it running after you log out from the server (then you should also add -u as option to python) and add an & to start it in the background. Also, you might want to direct the std:out and std:err into a file using > but all of that is just standard shell use and has nothing to do with the toolkit.

python run_training_pipeline.py <shorthand of the pipeline>

You can supply any of the following arguments, but don't have to (although for training you should definitely specify at least a GPU ID).

--gpu_id <ID of the GPU you wish to use, as displayed with nvidia-smi, default is cpu> 

--resume_checkpoint <path to a checkpoint to load>

--resume (if this is present, the furthest checkpoint available will be loaded automatically)

--finetune (if this is present, the provided checkpoint will be fine-tuned on the data from this pipeline)

--model_save_dir <path to a directory where the checkpoints should be saved>

After every epoch, some logs will be written to the console. If the loss becomes NaN, you'll need to use a smaller learning rate or more warmup steps in the arguments of the call to the training_loop in the pipeline you are running.

If you get cuda out of memory errors, you need to decrease the batchsize in the arguments of the call to the training_loop in the pipeline you are running. Try decreasing the batchsize in small steps until you get no more out of cuda memory errors. Decreasing the batchsize may also require you to use a smaller learning rate. The use of GroupNorm should make it so that the training remains mostly stable.

Speaking of plots: in the directory you specified for saving model's checkpoint files and self-explanatory visualization data will appear. Since the checkpoints are quite big, only the five most recent ones will be kept. Training will stop after 100,000 update steps have been made by default for Tacotron2, 300,000 for FastSpeech2, and after 500,000 steps for HiFiGAN. Depending on the machine and configuration you are using this will take between 2 and 4 days, so verify that everything works on small tests before running the big thing. If you want to stop earlier, just kill the process, since everything is daemonic all the child-processes should die with it.

After training is complete, it is recommended to run run_weight_averaging.py. If you made no changes to the architectures and stuck to the default directory layout, it will automatically load any models you produced with one pipeline, average their parameters to get a slightly more robust model and save the result as best.pt in the same directory where all the corresponding checkpoints lie. This also compresses the file size slightly, so you should do this and then use the best.pt model for inference.

To build a new InferenceInterface, which you can then use for super simple inference, we're going to use an existing one as template again. If you use multi-speaker, take the LibriTTS ones as template, otherwise take any other one. Make a copy of the InferenceInterface. Change the name of the class in the copy and change the paths to the models to use the trained models of your choice. Instantiate the model with the same hyperparameters that you used when you created it in the corresponding training pipeline. The last thing to check is the language that you supply to the text frontend. Make sure it matches what you used during training.

With your newly created InferenceInterface, you can use your trained models pretty much anywhere, e.g. in other projects. All you need is the Utility directory, the Layers directory, the Preprocessing directory and the InferenceInterfaces directory (and of course your model checkpoint). That's all the code you need, it works standalone.

An InferenceInterface contains 2 useful methods. They are read_to_file and read_aloud.

• read_to_file takes as input a list of strings and a filename. It will synthesize the sentences in the list and concatenate them with a short pause inbetween and write them to the filepath you supply as the other argument.

• read_aloud takes just a string, which it will then convert to speech and immediately play using the system's speakers. If you set the optional argument view to True when calling it, it will also show a plot of the phonemes it produced, the spectrogram it came up with, and the wave it created from that spectrogram. So all the representations can be seen, text to phoneme, phoneme to spectrogram and finally spectrogram to wave.

• Additionally, Tacotron2 InferenceInterfaces offer a method called plot_attention. This will take a string, synthesize it and show a plot of the attention matrix, which can be useful to gain insights.

Those methods are used in demo code in the toolkit. In run_interactive_demo.py and run_text_to_file_reader.py, you can import InferenceInterfaces that you created and add them to the dictionary in each of the files with a shorthand that makes sense. In the interactive demo, you can just call the python script, then type in the shorthand when prompted and immediately listen to your synthesis saying whatever you put in next (be wary of out of memory errors for too long inputs). In the text reader demo script you have to call the function that wraps around the InferenceInterface and supply the shorthand of your choice. It should be pretty clear from looking at it.

Here are a few points that were brought up by users:

• My error message shows GPU0, even though I specified a different GPU - The way GPU selection works is that the specified GPU is set as the only visible device, in order to avoid backend stuff running accidentally on different GPUs. So internally the program will name the device GPU0, because it is the only GPU it can see. It is actually running on the GPU you specified.
• I'm getting device side assert triggered errors - The amount of phonemes in the phoneme set used has to be specified as idim in the instantiation of a model. If a phoneme index is passed to the model during runtime which is higher than the amount specified as idim, this is the cryptic error that will occur. So if you make changes to the phoneme set, remember to also change the models idim.

This toolkit has been written by Florian Lux (except for the pytorch modules taken from ESPnet and ParallelWaveGAN, as mentioned above), so if you come across problems or questions, feel free to write a mail. Also let me know if you do something cool with it. Thank you for reading.

@inproceedings{lux2021toucan,
  title={{The IMS Toucan system for the Blizzard Challenge 2021}},
  author={Florian Lux and Julia Koch and Antje Schweitzer and Ngoc Thang Vu},
  year={2021},
  booktitle={Proc. Blizzard Challenge Workshop},
  volume={2021},
  publisher={{Speech Synthesis SIG}}
}