DeepLearningExamples/PyTorch/SpeechSynthesis/FastPitch

FastPitch 1.0 for PyTorch

This repository provides a script and recipe to train the FastPitch model to achieve state-of-the-art accuracy and is tested and maintained by NVIDIA.

Table Of Contents

• Model overview
  • Model architecture
  • Default configuration
  • Feature support matrix
    • Features
  • Mixed precision training
    • Enabling mixed precision
  • Glossary
• Setup
  • Requirements
• Quick Start Guide

• Advanced

  • Scripts and sample code
  • Parameters
  • Command-line options
  • Getting the data
    • Dataset guidelines
    • Multi-dataset
  • Training process
  • Inference process

• Performance

  • Benchmarking
    • Training performance benchmark
    • Inference performance benchmark
  • Results
    • Training accuracy results
      • Training accuracy: NVIDIA DGX-1 (8x V100 16G)
    • Training performance results
      • Training performance: NVIDIA DGX-1 (8x V100 16G)
      • Expected training time
    • Inference performance results
      • Inference performance: NVIDIA DGX-1 (1x V100 16G)
      • Inference performance: NVIDIA T4

• Release notes

  • Changelog
  • Known issues

Model overview

FastPitch is one of two major components in a neural, text-to-speech (TTS) system:

• a mel-spectrogram generator such as FastPitch or Tacotron 2, and
• a waveform synthesizer such as WaveGlow (see NVIDIA example code).

Such two-component TTS system is able to synthesize natural sounding speech from raw transcripts.

The FastPitch model generates mel-spectrograms and predicts a pitch contour from raw input text. It allows to exert additional control over the synthesized utterances, such as:

• modify the pitch contour to control the prosody,
• increase or decrease the fundamental frequency in a naturally sounding way, that preserves the perceived identity of the speaker,
• alter the pace of speech. The FastPitch model is based on the FastSpeech model. The main differences between FastPitch and FastSpeech are that FastPitch:
• explicitly learns to predict the pitch contour,
• pitch conditioning removes harsh sounding artifacts and provides faster convergence,
• no need for distilling mel-spectrograms with a teacher model,
• character durations are extracted with a pre-trained Tacotron 2 model.

FastPitch is trained on a publicly available LJ Speech dataset.

This model is trained with mixed precision using Tensor Cores on NVIDIA Volta and Turing GPUs. Therefore, researchers can get results from 2.0x to 2.7x2.2x faster than training without Tensor Cores, while experiencing the benefits of mixed precision training. This model is tested against each NGC monthly container release to ensure consistent accuracy and performance over time.

Model architecture

FastPitch is a fully feedforward Transformer model that predicts mel-spectrograms from raw text (Figure 1). The entire process is parallel, which means that all input letters are processed simultaneously to produce a full mel-spectrogram in a single forward pass.

Figure 1. Architecture of FastPitch. The model is composed of a bidirectional Transformer backbone (also known as a Transformer encoder), a pitch predictor, and a duration predictor. After passing through the first *N* Transformer blocks, encoding, the signal is augmented with pitch information and discretely upsampled. Then it goes through another set of *N* Transformer blocks, with the goal of smoothing out the upsampled signal, and constructing a mel-spectrogram.

Default configuration

The FastPitch model supports multi-GPU and mixed precision training with dynamic loss scaling (see Apex code here), as well as mixed precision inference.

The following features were implemented in this model:

• data-parallel multi-GPU training,
• dynamic loss scaling with backoff for Tensor Cores (mixed precision) training,
• gradient accumulation for reproducible results regardless of the number of GPUs.

To speed-up FastPitch training, reference mel-spectrograms, character durations, and pitch cues are generated during the pre-processing step and read directly from the disk during training. For more information on data pre-processing refer to Dataset guidelines and the paper.

Feature support matrix

The following features are supported by this model.

Feature	FastPitch
AMP	Yes
Apex DistributedDataParallel	Yes

Features

AMP - a tool that enables Tensor Core-accelerated training. For more information, refer to Enabling mixed precision.

Apex DistributedDataParallel - a module wrapper that enables easy multiprocess distributed data parallel training, similar to torch.nn.parallel.DistributedDataParallel. DistributedDataParallel is optimized for use with NCCL. It achieves high performance by overlapping communication with computation during backward() and bucketing smaller gradient transfers to reduce the total number of transfers required.

Mixed precision training

Mixed precision is the combined use of different numerical precisions in a computational method. Mixed precision training offers significant computational speedup by performing operations in half-precision format while storing minimal information in single-precision to retain as much information as possible in critical parts of the network. Since the introduction of Tensor Cores in the Volta and Turing architecture, significant training speedups are experienced by switching to mixed precision -- up to 3x overall speedup on the most arithmetically intense model architectures. Using mixed precision training requires two steps:

1. Porting the model to use the FP16 data type where appropriate.
   
2. Adding loss scaling to preserve small gradient values.

The ability to train deep learning networks with lower precision was introduced in the Pascal architecture and first supported in CUDA 8 in the NVIDIA Deep Learning SDK.

For information about:

• How to train using mixed precision, see the Mixed Precision Training paper and Training With Mixed Precision documentation.
• Techniques used for mixed precision training, see the Mixed-Precision Training of Deep Neural Networks blog.
• APEX tools for mixed precision training, see the NVIDIA Apex: Tools for Easy Mixed-Precision Training in PyTorch.

Enabling mixed precision

Mixed precision is enabled in PyTorch by using the Automatic Mixed Precision (AMP) library from APEX that casts variables to half-precision upon retrieval, while storing variables in single-precision format. Furthermore, to preserve small gradient magnitudes in backpropagation, a loss scaling step must be included when applying gradients. In PyTorch, loss scaling can be easily applied by using the scale_loss() method provided by AMP. The scaling value to be used can be dynamic or fixed.

By default, the train_fastpitch.sh script will launch mixed precision training with Tensor Cores. You can change this behaviour by removing the --amp-run flag from the train.py script.

To enable mixed precision, the following steps were performed:

• Import AMP from APEX:

  from apex import amp
  

• Initialize AMP:

  model, optimizer = amp.initialize(model, optimizer, opt_level="O1")
  

• If running on multi-GPU, wrap the model with DistributedDataParallel:

  from apex.parallel import DistributedDataParallel as DDP
  model = DDP(model)
  

• Scale loss before backpropagation (assuming loss is stored in a variable called losses):

  • Default backpropagate for FP32:

    losses.backward()
    

  • Scale loss and backpropagate with AMP:

    with optimizer.scale_loss(losses) as scaled_losses:
        scaled_losses.backward()
    

Glossary

Character duration The time during which a character is being articulated. Could be measured in milliseconds, mel-spectrogram frames, etc. Some characters are not pronounced, and thus have 0 duration.

Forced alignment Segmentation of a recording into lexical units like characters, words, or phonemes. The segmentation is hard and defines exact starting and ending times for every unit.

Fundamental frequency The lowest vibration frequency of a periodic soundwave, for example, produced by a vibrating instrument. It is perceived as the loudest. In the context of speech, it refers to the frequency of vibration of vocal chords. Abbreviated as f0.

Pitch A perceived frequency of vibration of music or sound.

Transformer
The paper Attention Is All You Need introduces a novel architecture called Transformer, which repeatedly applies the attention mechanism. It transforms one sequence into another.

Setup

The following section lists the requirements that you need to meet in order to start training the FastPitch model.

Requirements

This repository contains Dockerfile which extends the PyTorch NGC container and encapsulates some dependencies. Aside from these dependencies, ensure you have the following components:

• NVIDIA Docker
• PyTorch 20.03-py3 NGC container or newer
• NVIDIA Volta or Turing based GPU

For more information about how to get started with NGC containers, see the following sections from the NVIDIA GPU Cloud Documentation and the Deep Learning Documentation:

• Getting Started Using NVIDIA GPU Cloud
• Accessing And Pulling From The NGC Container Registry
• Running PyTorch
  

For those unable to use the PyTorch NGC container, to set up the required environment or create your own container, see the versioned NVIDIA Container Support Matrix.

Quick Start Guide

To train your model using mixed precision with Tensor Cores or using FP32, perform the following steps using the default parameters of the FastPitch model on the LJSpeech 1.1 dataset. For the specifics concerning training and inference, see the Advanced section.

1. Clone the repository.

   git clone https://github.com/NVIDIA/DeepLearningExamples.git
   cd DeepLearningExamples/PyTorch/SpeechSynthesis/FastPitch
   

2. Build and run the FastPitch PyTorch NGC container.

By default the container will use the first available GPU. Modify the script to include other available devices.

   bash scripts/docker/build.sh
   bash scripts/docker/interactive.sh

1. Download and preprocess the dataset.

Use the scripts to automatically download and preprocess the training, validation and test datasets:

   bash scripts/download_dataset.sh
   bash scripts/prepare_dataset.sh

The data is downloaded to the ./LJSpeech-1.1 directory (on the host). The ./LJSpeech-1.1 directory is mounted under the /workspace/fastpitch/LJSpeech-1.1 location in the NGC container. The complete dataset has the following structure:

   ./LJSpeech-1.1
   ├── durations        # Character durations estimates for forced alignment training
   ├── mels             # Pre-calculated target mel-spectrograms
   ├── metadata.csv     # Mapping of waveforms to utterances
   ├── pitch_char       # Average per-character fundamental frequencies for input utterances
   ├── pitch_char_stats__ljs_audio_text_train_filelist.json    # Mean and std of pitch for training data
   ├── README
   └── wavs             # Raw waveforms

1. Start training.

   bash scripts/train.sh
   

   The training will produce a FastPitch model capable of generating mel-spectrograms from raw text. It will be serialized as a single .pt checkpoint file, along with a series of intermediate checkpoints. The script is configured for 8x GPU with 16GB of memory. Consult Training process and example configs to adjust to a different configuration.

2. Start validation/evaluation.

Ensure your training loss values are comparable to those listed in the table in the Results section. Note that the validation loss is evaluated with ground truth durations for letters (not the predicted ones). The loss values are stored in the ./output/nvlog.json log file, ./output/{train,val,test} as TensorBoard logs, and printed to the standard output (stdout) during training. The main reported loss is a weighted sum of losses for mel-, pitch-, and duration- predicting modules.

The audio can be generated by following the Inference process section below. The synthesized audio should be similar to the samples in the ./audio directory.

1. Start inference/predictions.

To synthesize audio, you will need a WaveGlow model, which generates waveforms based on mel-spectrograms generated with FastPitch. By now, a pre-trained model should have been downloaded by the scripts/download_dataset.sh script. Alternatively, to train WaveGlow from scratch, follow the instructions in NVIDIA/DeepLearningExamples/Tacotron2 and replace the checkpoint in the ./pretrained_models/waveglow directory.

You can perform inference using the respective .pt checkpoints that are passed as --fastpitch and --waveglow arguments:

   python inference.py --cuda --wn-channels 256 --amp-run \
                       --fastpitch output/<FastPitch checkpoint> \
                       --waveglow pretrained_models/waveglow/<waveglow checkpoint> \
                       -i phrases/devset10.tsv \
                       -o output/wavs_devset10

The speech is generated from a file passed with the -i argument, with one utterance per line:

   `<output wav file name>|<utterance>`

To run inference in mixed precision, use the --amp-run flag. The output audio will be stored in the path specified by the -o argument. Consult the inference.py to learn more options, such as setting the batch size.

Advanced

The following sections provide greater details of the dataset, running training and inference, and the training results.

Scripts and sample code

The repository holds code for FastPitch (training and inference) and WaveGlow (inference only). The code specific to a particular model is located in that model’s directory - ./fastpitch and ./waveglow - and common functions live in the ./common directory. The model-specific scripts are as follows:

• <model_name>/model.py - the model architecture, definition of forward and inference functions
• <model_name>/arg_parser.py - argument parser for parameters specific to a given model
• <model_name>/data_function.py - data loading functions
• <model_name>/loss_function.py - loss function for the model

The common scripts contain layer definitions common to both models (common/layers.py), some utility scripts (common/utils.py) and scripts for audio processing (common/audio_processing.py and common/stft.py).

In the root directory ./ of this repository, the ./train.py script is used for training while inference can be executed with the ./inference.py script. The scripts ./models.py, ./data_functions.py and ./loss_functions.py call the respective scripts in the <model_name> directory, depending on what the model is trained using the train.py script.

The repository is structured similarly to the NVIDIA Tacotron2 Deep Learning example, so that they could be combined in more advanced use cases.

Parameters

In this section, we list the most important hyperparameters and command-line arguments, together with their default values that are used to train FastPitch.

• --epochs - number of epochs (default: 1500)
• --learning-rate - learning rate (default: 0.1)
• --batch-size - batch size (default: 32)

• --amp-run - use mixed precision training

• --pitch-predictor-loss-scale - rescale the loss of the pitch predictor module to dampen its influence on the shared feedforward transformer blocks

• --duration-predictor-loss-scale - rescale the loss of the duration predictor module to dampen its influence on the shared feedforward transformer blocks

Command-line options

To see the full list of available options and their descriptions, use the -h or --help command line option, for example:

python train.py --help

The following example output is printed when running the model:

DLL 2020-03-30 10:41:12.562594 - epoch    1 | iter   1/19 | loss 36.99 | mel loss 35.25 |  142370.52 items/s | elapsed 2.50 s | lrate 1.00E-01 -> 3.16E-06
DLL 2020-03-30 10:41:13.202835 - epoch    1 | iter   2/19 | loss 37.26 | mel loss 35.98 |  561459.27 items/s | elapsed 0.64 s | lrate 3.16E-06 -> 6.32E-06
DLL 2020-03-30 10:41:13.831189 - epoch    1 | iter   3/19 | loss 36.93 | mel loss 35.41 |  583530.16 items/s | elapsed 0.63 s | lrate 6.32E-06 -> 9.49E-06

Getting the data

The FastPitch and WaveGlow models were trained on the LJSpeech-1.1 dataset. The ./scripts/download_dataset.sh script will automatically download and extract the dataset to the ./LJSpeech-1.1 directory.

Dataset guidelines

The LJSpeech dataset has 13,100 clips that amount to about 24 hours of speech of a single, female speaker. Since the original dataset does not define a train/dev/test split of the data, we provide a split in the form of three file lists:

./filelists
├── ljs_mel_dur_pitch_text_test_filelist.txt
├── ljs_mel_dur_pitch_text_train_filelist.txt
└── ljs_mel_dur_pitch_text_val_filelist.txt

NOTE: When combining FastPitch/WaveGlow with external models trained on LJSpeech-1.1, make sure that your train/dev/test split matches. Different organizations may use custom splits. A mismatch poses a risk of leaking the training data through model weights during validation and testing.

FastPitch predicts character durations just like FastSpeech does. This calls for training with forced alignments, expressed as the number of output mel-spectrogram frames for every input character. To this end, a pre-trained Tacotron 2 model is used. Its attention matrix relates the input characters with the output mel-spectrogram frames.

For every mel-spectrogram frame, its fundamental frequency in Hz is estimated with Praat. Those values are then averaged over every character, in order to provide sparse pitch cues for the model. Character boundaries are calculated from durations extracted previously with Tacotron 2.

Figure 2. Pitch estimates for mel-spectrogram frames of phrase "in being comparatively" (in blue) averaged over characters (in red). Silent letters have duration 0 and are omitted.

Multi-dataset

Follow these steps to use datasets different from the default LJSpeech dataset.

1. Prepare a directory with .wav files.

   ./my_dataset
   └── wavs
   

2. Prepare filelists with transcripts and paths to .wav files. They define training/validation split of the data (test is currently unused):

   ./filelists
   ├── my_dataset_mel_ali_pitch_text_train_filelist.txt
   └── my_dataset_mel_ali_pitch_text_val_filelist.txt
   

Those filelists should list a single utterance per line as:

   `<audio file path>|<transcript>`

The <audio file path> is the relative path to the path provided by the --dataset-path option of train.py.

1. Run the pre-processing script to calculate mel-spectrograms, durations and pitch:

   python extract_mels.py --cuda \
                         --dataset-path ./my_dataset \
                         --wav-text-filelist ./filelists/my_dataset_mel_ali_pitch_text_train_filelist.txt \
                         --extract-mels \
                         --extract-durations \
                         --extract-pitch-char \
                         --tacotron2-checkpoint ./pretrained_models/tacotron2/state_dict.pt"
   
   python extract_mels.py --cuda \
                         --dataset-path ./my_dataset \
                         --wav-text-filelist ./filelists/my_dataset_mel_ali_pitch_text_val_filelist.txt \
                         --extract-mels \
                         --extract-durations \
                         --extract-pitch-char \
                         --tacotron2-checkpoint ./pretrained_models/tacotron2/state_dict.pt"
   

In order to use the prepared dataset, pass the following to the train.py script:

   --dataset-path ./my_dataset` \
   --training-files ./filelists/my_dataset_mel_ali_pitch_text_train_filelist.txt \
   --validation files ./filelists/my_dataset_mel_ali_pitch_text_val_filelist.txt

Training process

FastPitch is trained to generate mel-spectrograms from raw text input. It uses short time Fourier transform (STFT) to generate target mel-spectrograms from audio waveforms to be the training targets.

The training loss is averaged over an entire training epoch, whereas the validation loss is averaged over the validation dataset. Performance is reported in total output mel-spectrogram frames per second and recorded as train_frames/s (after each iteration) and avg_train_frames/s (averaged over epoch) in the output log file ./output/nvlog.json. The result is averaged over an entire training epoch and summed over all GPUs that were included in the training.

The scripts/train.sh script is configured for 8x GPU with 16GB of memory:

```
--batch-size 32
--gradient-accumulation-steps 1
```

In a single accumulated step, there are batch_size x gradient_accumulation_steps x #GPUs = 256 examples being processed in parallel. With a smaller number of GPUs, increase --gradient_accumulation_steps to keep this relation satisfied.

Even though the training script uses all available GPUs, you can select the devices with the CUDA_VISIBLE_DEVICES environmental variable (see this CUDA Pro Tip for more details).

Inference process

You can run inference using the ./inference.py script. This script takes text as input and runs FastPitch and then WaveGlow inference to produce an audio file. It requires pre-trained checkpoints of both models and input text as a text file, with one phrase per line.

Having pre-trained models in place, run the sample inference on LJSpeech-1.1 test-set with:

bash scripts/inference_example.sh

Examine the inference_example.sh script to adjust paths to pre-trained models, and call python inference.py --help to learn all available options. By default, synthesized audio samples are saved in ./output/audio_* folders. The audio files sample_fp16.wav and sample_fp32.wav were generated using checkpoints from mixed precision and FP32 training, respectively.

FastPitch allows us to linearly adjust the pace of synthesized speech like FastSpeech. For instance, pass --pace 0.5 for a twofold decrease in speed.

For every input character, the model predicts a pitch cue - an average pitch over a character in Hz. Pitch can be adjusted by transforming those pitch cues. A few simple examples are provided below.

The flags can be combined. Modify these functions directly in the inference.py script to gain more control over the final result.

You can find all the available options by calling python inference.py --help.

Performance

Benchmarking

The following section shows how to run benchmarks measuring the model performance in training and inference mode.

Training performance benchmark

To benchmark the training performance on a specific batch size, run:

FastPitch

• For 1 GPU

  • FP16

    python train.py \
        --amp-run \
        --batch-size 64 \
        --gradient-accumulation-steps 4 \
        --cuda \
        --cudnn-enabled \
        -o <output-dir> \
        --log-file <output-dir>/nvlog.json \
        --dataset-path <dataset-path> \
        --training-files <train-filelist-path> \
        --validation-files <val-filelist-path> \
        --pitch-mean-std-file <pitch-stats-path> \
        --epochs 10 \
        --warmup-steps 1000 \
        -lr 0.1 \
        --optimizer lamb \
        --gad-clip-thresh 1000.0 \
        --dur-predictor-loss-scale 0.1 \
        --pitch-predictor-loss-scale 0.1 \
        --weight-decay 1e-6
    

  • FP32

    python train.py \
        --batch-size 32 \
        --gradient-accumulation-steps 8 \
        --cuda \
        --cudnn-enabled \
        -o <output-dir> \
        --log-file <output-dir>/nvlog.json \
        --dataset-path <dataset-path> \
        --training-files <train-filelist-path> \
        --validation-files <val-filelist-path> \
        --pitch-mean-std-file <pitch-stats-path> \
        --epochs 10 \
        --warmup-steps 1000 \
        -lr 0.1 \
        --optimizer lamb \
        --grad-clip-thresh 1000.0 \
        --dur-predictor-loss-scale 0.1 \
        --pitch-predictor-loss-scale 0.1 \
        --weight-decay 1e-6
    

• For multiple GPUs

  • FP16

    python -m multiproc train.py \
        --amp-run \
        --batch-size 32 \
        --gradient-accumulation-steps 1 \
        --cuda \
        --cudnn-enabled \
        -o <output-dir> \
        --log-file <output-dir>/nvlog.json \
        --dataset-path <dataset-path> \
        --training-files <train-filelist-path> \
        --validation-files <val-filelist-path> \
        --pitch-mean-std-file <pitch-stats-path> \
        --epochs 10 \
        --warmup-steps 1000 \
        -lr 0.1 \
        --optimizer lamb \
        --grad-clip-thresh 1000.0 \
        --dur-predictor-loss-scale 0.1 \
        --pitch-predictor-loss-scale 0.1 \
        --weight-decay 1e-6
    

  • FP32

    python -m multiproc train.py \
        --batch-size 32 \
        --gradient-accumulation-steps 1
        --cuda \
        --cudnn-enabled \
        -o <output-dir> \
        --log-file <output-dir>/nvlog.json \
        --dataset-path <dataset-path> \
        --training-files <train-filelist-path> \
        --validation-files <val-filelist-path> \
        --pitch-mean-std-file <pitch-stats-path> \
        --epochs 10 \
        --warmup-steps 1000 \
        -lr 0.1 \
        --optimizer lamb \
        --grad-clip-thresh 1000.0 \
        --dur-predictor-loss-scale 0.1 \
        --pitch-predictor-loss-scale 0.1 \
        --weight-decay 1e-6
    

Each of these scripts runs for 10 epochs and for each epoch measures the average number of items per second. The performance results can be read from the nvlog.json files produced by the commands.

Inference performance benchmark

To benchmark the inference performance on a specific batch size, run:

• For FP16

  python inference.py --cuda --amp-run \
                       --fastpitch output/checkpoint_FastPitch_1500.pt \
                       --waveglow pretrained_models/waveglow/waveglow_256channels_ljs_v3.pt \
                       --wn-channels 256 \
                       --include-warmup \
                       --batch-size 1 \
                       --repeats 1000 \
                       --input phrases/benchmark_8_128.tsv \
                       --log-file output/nvlog_inference.json
  

• For FP32

  python inference.py --cuda \
                       --fastpitch output/checkpoint_FastPitch_1500.pt \
                       --waveglow pretrained_models/waveglow/waveglow_256channels_ljs_v3.pt \
                       --wn-channels 256 \
                       --include-warmup \
                       --batch-size 1 \
                       --pitch \
                       --repeats 1000 \
                       --input phrases/benchmark_8_128.tsv \
                       --log-file output/nvlog_inference.json
  

The output log files will contain performance numbers for the FastPitch model (number of output mel-spectrogram frames per second, reported as generator_frames/s) and for WaveGlow (number of output samples per second, reported as waveglow_samples/s). The inference.py script will run a few warm-up iterations before running the benchmark. Inference will be averaged over 100 runs, as set by the --repeats flag.

Results

The following sections provide details on how we achieved our performance and accuracy in training and inference.

Training accuracy results

Training accuracy: NVIDIA DGX-1 (8x V100 16G)

Our results were obtained by running the ./platform/train_fastpitch_{AMP,FP32}_DGX1_16GB_8GPU.sh training script in the PyTorch 20.03-py3 NGC container on NVIDIA DGX-1 with 8x V100 16G GPUs.

All of the results were produced using the train.py script as described in the Training process section of this document.

Transformation	Flag	Samples
Amplify pitch wrt. to the mean pitch	--pitch-transform-amplify	link
Invert pitch wrt. to the mean pitch	--pitch-transform-invert	link
Raise/lower pitch by	--pitch-transform-shift <hz>	link
Flatten the pitch to a constant value	--pitch-transform-flatten	link
Change the pace of speech (1.0 = unchanged)	--pace <value>	link

Loss (Model/Epoch)	0	250	500	750	1000	1250	1500
FastPitch AMP	35.094	0.254	0.216	0.201	0.193	0.187	0.184
FastPitch FP32	35.108	0.254	0.216	0.200	0.194	0.188	0.184

Training performance results

Training performance: NVIDIA DGX-1 (8x V100 16G)

Our results were obtained by running the ./platform/train_fastpitch_{AMP,FP32}_DGX1_16GB_8GPU.sh training script in the PyTorch 20.03-py3 NGC container on NVIDIA DGX-1 with 8x V100 16G GPUs. Performance numbers, in output mel-spectrograms per second, were averaged over an entire training epoch.

Number of GPUs	Batch size per GPU	Number of mels used with AMP	Number of mels used with FP32	Speed-up with AMP	Multi-GPU strong scaling with AMP	Multi-GPU strong scaling with FP32
1	64@AMP, 32@FP32	109769	40636	2.70	1.00	1.00
4	64@AMP, 32@FP32	361195	150921	2.39	3.29	3.71
8	32@AMP, 32@FP32	562136	278778	2.02	5.12	6.86

To achieve these same results, follow the steps in the Quick Start Guide.

Expected training time

The following table shows the expected training time for convergence for 1500 epochs:

Number of GPUs	Batch size per GPU	Time to train with AMP (Hrs)	Time to train with FP32 (Hrs)	Speed-up with AMP
1	64@AMP, 32@FP32	27.0	73.7	2.73
4	64@AMP, 32@FP32	8.4	19.7	2.36
8	32@AMP, 32@FP32	5.5	10.8	1.97

Note that most of the quality is achieved after the initial 500 epochs.

Inference performance results

The following tables show inference statistics for the FastPitch and WaveGlow text-to-speech system, gathered from 1000 inference runs, on a single V100 and a single T4, respectively. Latency is measured from the start of FastPitch inference to the end of WaveGlow inference. Throughput is measured as the number of generated audio samples per second at 22KHz. RTF is the real-time factor which denotes the number of seconds of speech generated in a second of wall-clock time, per input utterance. The used WaveGlow model is a 256-channel model published on NGC.

Our results were obtained by running the ./scripts/inference_benchmark.sh script in the PyTorch 20.03-py3 NGC container. Note that to reproduce the results, you need to provide pre-trained checkpoins for FastPitch and WaveGlow. Edit the script to provide your checkpoint filenames.

Note that performance numbers are related to the length of input. The numbers reported below were taken with a moderate length of 128 characters. For longer utterances even better numbers are expected, as the generator is fully parallel.

Inference performance: NVIDIA DGX-1 (1x V100 16G)

The input utterance has 128 characters, synthesized audio has 8.05 s.

Batch size	Precision	Avg latency (s)	Latency tolerance interval 90% (s)	Latency tolerance interval 95% (s)	Latency tolerance interval 99% (s)	Throughput (samples/sec)	Speed-up with mixed precision	Avg RTF
1	FP16	0.253	0.254	0.255	0.255	702,735	1.51	31.87
4	FP16	0.572	0.575	0.575	0.576	1,243,094	2.55	14.09
8	FP16	1.118	1.121	1.121	1.123	1,269,479	2.70	7.20
1	FP32	0.382	0.384	0.384	0.385	464,920	-	21.08
4	FP32	1.458	1.461	1.461	1.462	486,756	-	5.52
8	FP32	3.015	3.023	3.024	3.027	470,741	-	2.67

Inference performance: NVIDIA T4

The input utterance has 128 characters, synthesized audio has 8.05 s.

Batch size	Precision	Avg latency (s)	Latency tolerance interval 90% (s)	Latency tolerance interval 95% (s)	Latency tolerance interval 99% (s)	Throughput (samples/sec)	Speed-up with mixed precision	Avg RTF
1	FP16	0.952	0.958	0.960	0.962	186,349	1.30	8.45
4	FP16	4.187	4.209	4.213	4.221	169,473	1.21	1.92
8	FP16	7.799	7.824	7.829	7.839	181,978	1.38	1.03
1	FP32	1.238	1.245	1.247	1.250	143,292	-	6.50
4	FP32	5.083	5.109	5.114	5.124	139,613	-	1.58
8	FP32	10.756	10.797	10.805	10.820	131,951	-	0.75

Release notes

Changelog

May 2020

• Initial release

Known issues

There are no known issues with this model with this model.