Tiny CUDA Neural Networks

This is a small, self-contained framework for training and querying neural networks. Most notably, it contains a lightning fast "fully fused" multi-layer perceptron as well as support for various advanced input encodings, losses, and optimizers.

License and Citation

This framework is licensed under the BSD 3-clause license. Please see LICENSE.txt for details.

If you use it in your research, we would appreciate a citation via

@misc{tiny-cuda-nn,
    Author = {Thomas M\"uller},
    Year = {2021},
    Note = {https://github.com/nvlabs/tiny-cuda-nn},
    Title = {Tiny {CUDA} Neural Network Framework}
}

For business inquiries, please visit our website and submit the form: NVIDIA Research Licensing

Performance

Fully fused networks vs. TensorFlow v2.5.0 w/ XLA. Measured on 64 (solid line) and 128 (dashed line) neurons wide multi-layer perceptrons on an RTX 3090. Generated by benchmarks/bench_ours.cu and benchmarks/bench_tensorflow.py.

Publications

This framework powers the following publications:

Instant Neural Graphics Primitives with a Multiresolution Hash Encoding
Thomas Müller, Alex Evans, Christoph Schied, Alexander Keller
arXiv [cs.GR], Jan 2022
[ Website ] [ Paper ] [ Code ] [ Video ]

Real-time Neural Radiance Caching for Path Tracing
Thomas Müller, Fabrice Rousselle, Jan Novák, Alexander Keller
ACM Transactions on Graphics (Proceedings of SIGGRAPH), vol. 40, no. 4, pp. 36:1–36:16, Aug 2021
[ Paper ] [ GTC talk ] [ Video ] [ Interactive Results Viewer ] [ BibTeX ]

Extracting Triangular 3D Models, Materials, and Lighting From Images
Jacob Munkberg, Jon Hasselgren, Tianchang Shen, Jun Gao, Wenzheng Chen, Alex Evans, Thomas Müller, Sanja Fidler
arXiv:2111.12503 [cs.CV], Nov 2021
[ Website ] [ Paper ] [ Video ] [ BibTeX ]

Usage

Tiny CUDA neural networks have a simple C++/CUDA API:

#include <tiny-cuda-nn/common.h>

// Configure the model
nlohmann::json config = {
	{"loss", {
		{"otype", "L2"}
	}},
	{"optimizer", {
		{"otype", "Adam"},
		{"learning_rate", 1e-3},
	}},
	{"encoding", {
		{"otype", "OneBlob"},
		{"n_bins", 32},
	}},
	{"network", {
		{"otype", "FullyFusedMLP"},
		{"n_neurons", 64},
		{"n_hidden_layers", 5},
		{"activation", "ReLU"},
		{"output_activation", "None"},
	}},
};

using namespace tcnn;

auto model = create_from_config(n_input_dims, n_output_dims, config);

// Train the model
GPUMatrix<float> training_batch_inputs(n_input_dims, batch_size);
GPUMatrix<float> training_batch_targets(n_output_dims, batch_size);

for (int i = 0; i < n_training_steps; ++i) {
	generate_training_batch(&training_batch_inputs, &training_batch_targets); // <-- your code

	float loss;
	model.trainer->training_step(training_batch_inputs, training_batch_targets, &loss);
	std::cout << "iteration=" << i << " loss=" << loss << std::endl;
}

// Use the model
GPUMatrix<float> inference_inputs(n_input_dims, batch_size);
generate_inputs(&inference_inputs); // <-- your code

GPUMatrix<float> inference_outputs(n_output_dims, batch_size);
model.network->inference(inference_inputs, inference_outputs);

Example: learning a 2D image

We provide a sample application where an image function (x,y) -> (R,G,B) is learned. It can be run via

tiny-cuda-nn/build> ./mlp_learning_an_image ../data/images/albert.exr ../data/config.json

producing an image every 1000 training steps. Each 1000 steps should take roughly 0.8 seconds with the default configuration on an RTX 3090.

Learned image after 1,000 steps	Learned image after 10,000 steps	Reference image

Requirements

• CUDA v10.2 or higher.
• CMake v3.18 or higher.
• A C++14 capable compiler.
• A high-end NVIDIA GPU that supports TensorCores and has a large amount of shared memory. The framework was tested primarily with an RTX 3090.
• The fully fused MLP component of this framework requires a very large amount of shared memory in its default configuration. It will likely only work on an RTX 3090, an RTX 2080 Ti, or high-end enterprise GPUs. Lower end cards must reduce the n_neurons parameter or use the CutlassMLP (better compatibility but slower) instead.

Compilation

Begin by cloning this repository and all its submodules using the following command:

$ git clone --recursive https://github.com/nvlabs/tiny-cuda-nn
$ cd tiny-cuda-nn

Then, use CMake to generate build files:

tiny-cuda-nn$ mkdir build
tiny-cuda-nn$ cd build
tiny-cuda-nn/build$ cmake ..

The last step differs by operating system.

• Windows: open tiny-cuda-nn/build/tiny-cuda-nn.sln in Visual Studio and click the "Build" button.
• Linux: run the command

  tiny-cuda-nn/build$ make -j

Components

The following is a summary of all components of this framework that are currently released. Please consult the JSON documentation for how to configure them.

Networks
Fully fused MLP	src/fully_fused_mlp.cu	Lightning fast implementation of small multi-layer perceptrons (MLPs).
CUTLASS MLP	src/cutlass_mlp.cu	MLP based on CUTLASS' GEMM routines. Slower than fully-fused, but handles larger networks and still is reasonably fast.
CUTLASS ResNet	src/cutlass_resnet.cu	Fully connected residual network based on CUTLASS' GEMM routines.

Input encodings
Composite	include/tiny-cuda-nn/encodings/composite.h	Allows composing multiple encodings. Can be, for example, used to assemble the Neural Radiance Caching encoding [Müller et al. 2021].
Frequency	include/tiny-cuda-nn/encodings/frequency.h	NeRF's [Mildenhall et al. 2020] positional encoding applied equally to all dimensions.
Grid	include/tiny-cuda-nn/encodings/grid.h	Encoding based on trainable multiresolution grids. Used for Instant Neural Graphics Primitives [Müller et al. 2022]. The grids can be backed by hashtables, dense storage, or tiled storage.
Identity	include/tiny-cuda-nn/encodings/identity.h	Leaves values untouched.
Oneblob	include/tiny-cuda-nn/encodings/oneblob.h	From Neural Importance Sampling [Müller et al. 2019] and Neural Control Variates [Müller et al. 2020].
SphericalHarmonics	include/tiny-cuda-nn/encodings/spherical_harmonics.h	A frequency-space encoding that is more suitable to direction vectors than component-wise ones.
TriangleWave	include/tiny-cuda-nn/encodings/triangle_wave.h	Low-cost alternative to the NeRF's encoding. Used in Neural Radiance Caching [Müller et al. 2021].

Losses
L1	include/tiny-cuda-nn/losses/l1.h	Standard L1 loss.
Relative L1	include/tiny-cuda-nn/losses/l1.h	Relative L1 loss normalized by the network prediction.
MAPE	include/tiny-cuda-nn/losses/mape.h	Mean absolute percentage error (MAPE). The same as Relative L1, but normalized by the target.
SMAPE	include/tiny-cuda-nn/losses/smape.h	Symmetric mean absolute percentage error (SMAPE). The same as Relative L1, but normalized by the mean of the prediction and the target.
L2	include/tiny-cuda-nn/losses/l2.h	Standard L2 loss.
Relative L2	include/tiny-cuda-nn/losses/relative_l2.h	Relative L2 loss normalized by the network prediction [Lehtinen et al. 2018].
Relative L2 Luminance	include/tiny-cuda-nn/losses/relative_l2_luminance.h	Same as above, but normalized by the luminance of the network prediction. Only applicable when network prediction is RGB. Used in Neural Radiance Caching [Müller et al. 2021].
Cross Entropy	include/tiny-cuda-nn/losses/cross_entropy.h	Standard cross entropy loss. Only applicable when the network prediction is a PDF.
Variance	include/tiny-cuda-nn/losses/variance_is.h	Standard variance loss. Only applicable when the network prediction is a PDF.

Optimizers
Adam	include/tiny-cuda-nn/optimizers/adam.h	Implementation of Adam [Kingma and Ba 2014], generalized to AdaBound [Luo et al. 2019].
Novograd	include/tiny-cuda-nn/optimizers/lookahead.h	Implementation of Novograd [Ginsburg et al. 2019].
SGD	include/tiny-cuda-nn/optimizers/sgd.h	Standard stochastic gradient descent (SGD).
Shampoo	include/tiny-cuda-nn/optimizers/shampoo.h	Implementation of the 2nd order Shampoo optimizer [Gupta et al. 2018] with home-grown optimizations as well as those by Anil et al. [2020].
Average	include/tiny-cuda-nn/optimizers/average.h	Wraps another optimizer and computes a linear average of the weights over the last N iterations. The average is used for inference only (does not feed back into training).
Batched	include/tiny-cuda-nn/optimizers/batched.h	Wraps another optimizer, invoking the nested optimizer once every N steps on the averaged gradient. Has the same effect as increasing the batch size but requires only a constant amount of memory.
EMA	include/tiny-cuda-nn/optimizers/average.h	Wraps another optimizer and computes an exponential moving average of the weights. The average is used for inference only (does not feed back into training).
Exponential Decay	include/tiny-cuda-nn/optimizers/exponential_decay.h	Wraps another optimizer and performs piecewise-constant exponential learning-rate decay.
Lookahead	include/tiny-cuda-nn/optimizers/lookahead.h	Wraps another optimizer, implementing the lookahead algorithm [Zhang et al. 2019].

Acknowledgments

Special thanks go to the NRC authors for helpful discussions and to Nikolaus Binder for providing part of the infrastructure of this framework, as well as for help with utilizing TensorCores from within CUDA.