The UNet model is a convolutional neural network for 2D image segmentation. This repository contains a UNet implementation as described in the original paper UNet: Convolutional Networks for Biomedical Image Segmentation, without any alteration.

This model is trained with mixed precision using Tensor Cores on Volta, Turing, and the NVIDIA Ampere GPU architectures. Therefore, researchers can get results 2.2x 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

UNet was first introduced by Olaf Ronneberger, Philip Fischer, and Thomas Brox in the paper: UNet: Convolutional Networks for Biomedical Image Segmentation. UNet allows for seamless segmentation of 2D images, with high accuracy and performance, and can be adapted to solve many different segmentation problems.

The following figure shows the construction of the UNet model and its different components. UNet is composed of a contractive and an expanding path, that aims at building a bottleneck in its centermost part through a combination of convolution and pooling operations. After this bottleneck, the image is reconstructed through a combination of convolutions and upsampling. Skip connections are added with the goal of helping the backward flow of gradients in order to improve the training.

Figure 1. The architecture of a UNet model. Taken from the UNet: Convolutional Networks for Biomedical Image Segmentation paper.

Default configuration

UNet consists of a contractive (left-side) and expanding (right-side) path. It repeatedly applies unpadded convolutions followed by max pooling for downsampling. Every step in the expanding path consists of an upsampling of the feature maps and concatenation with the correspondingly cropped feature map from the contractive path.

Feature support matrix

The following features are supported by this model:

Features

Automatic Mixed Precision (AMP)

This implementation of UNet uses AMP to implement mixed precision training. It allows us to use FP16 training with FP32 master weights by modifying just a few lines of code.

Horovod

Horovod is a distributed training framework for TensorFlow, Keras, PyTorch, and MXNet. The goal of Horovod is to make distributed deep learning fast and easy to use. For more information about how to get started with Horovod, see the Horovod: Official repository.

Multi-GPU training with Horovod

Our model uses Horovod to implement efficient multi-GPU training with NCCL. For details, see example sources in this repository or see the TensorFlow tutorial.

XLA support (experimental)

XLA is a domain-specific compiler for linear algebra that can accelerate TensorFlow models with potentially no source code changes. The results are improvements in speed and memory usage: most internal benchmarks run ~1.1-1.5x faster after XLA is enabled.

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 Volta, and following with both the Turing and Ampere architectures, 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 previously required two steps:

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

This can now be achieved using Automatic Mixed Precision (AMP) for TensorFlow to enable the full mixed precision methodology in your existing TensorFlow model code. AMP enables mixed precision training on Volta and Turing GPUs automatically. The TensorFlow framework code makes all necessary model changes internally.

In TF-AMP, the computational graph is optimized to use as few casts as necessary and maximize the use of FP16, and the loss scaling is automatically applied inside of supported optimizers. AMP can be configured to work with the existing tf.contrib loss scaling manager by disabling the AMP scaling with a single environment variable to perform only the automatic mixed-precision optimization. It accomplishes this by automatically rewriting all computation graphs with the necessary operations to enable mixed precision training and automatic loss scaling.

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.
• How to access and enable AMP for TensorFlow, see Using TF-AMP from the TensorFlow User Guide.

Enabling mixed precision

This implementation exploits the TensorFlow Automatic Mixed Precision feature. To enable AMP, you simply need to supply the --amp flag to the main.py script. For reference, enabling the AMP required us to apply the following changes to the code:

1. Set Keras mixed precision policy:

   if params['use_amp']:
       tf.keras.mixed_precision.experimental.set_policy('mixed_float16')
   

2. Use loss scaling wrapper on the optimizer:

   optimizer = tf.keras.optimizers.Adam(learning_rate=learning_rate)
   if using_amp:
       optimizer = tf.keras.mixed_precision.LossScaleOptimizer(optimizer, dynamic=True)
   

3. Use scaled loss to calculate gradients:

   scaled_loss = optimizer.get_scaled_loss(loss)
   tape = hvd.DistributedGradientTape(tape)
   scaled_gradients = tape.gradient(scaled_loss, model.trainable_variables)
   gradients = optimizer.get_unscaled_gradients(scaled_gradients)
   optimizer.apply_gradients(zip(gradients, model.trainable_variables))
   

Enabling TF32

TensorFloat-32 (TF32) is the new math mode in NVIDIA A100 GPUs for handling the matrix math also called tensor operations. TF32 running on Tensor Cores in A100 GPUs can provide up to 10x speedups compared to single-precision floating-point math (FP32) on Volta GPUs.

TF32 Tensor Cores can speed up networks using FP32, typically with no loss of accuracy. It is more robust than FP16 for models which require high dynamic range for weights or activations.

For more information, refer to the TensorFloat-32 in the A100 GPU Accelerates AI Training, HPC up to 20x blog post.

TF32 is supported in the NVIDIA Ampere GPU architecture and is enabled by default.