Pytorch

4.1 PyTorch can’t actually do big linear algebra calcs

Received an error like this while trying to do svd or eigh on a big matrix?

pytorch RuntimeError: false INTERNAL ASSERT FAILED please report a bug to PyTorch.
linalg.eigh: Argument 8 has illegal value.
    Most certainly there is a bug in the implementation calling the backend library.

Me too. We are screwed. Pytorch can’t actually do big linear algebra calcs. Possibly an experimental custom build might help? See #51720 · pytorch/pytorch.

4.2 Mutation

To mutate the content of a tensor, use the (to my mind non-obvious) copy_ method.

4.3 Complex numbers

Complex numbers work but complex parameters are still fragile.

4.4 do not use torch.diag

torch.diag ambiguously returns a diagonal matrix from a vector or extracts a diagonal from a matrix. You’d think there is never confusion between these? But in the sloppily-typed world of PyTorch tensors, there can be much ambiguity. This ambiguity just cost me 2 days of my life. There are two alternatives that do what I want, without introducing ambiguity:

torch.diagonal() always returns the vector of the diagonal elements of the input.

torch.diagflat() always constructs a tensor with diagonal elements specified by the input.

5.1 On Apple devices

MPS backend is supported in torch 2.0 onwards. See the worked example by Thai Tran.

6.1 Hooks

• Hooks: the one PyTorch trick you must know

6.2 Pretty printing

Lovely Tensors pretty-prints PyTorch tensors in a manner more informative than the default display.

Was it really useful for you, as a human, to see all these numbers?

What is the shape? The size? What are the statistics? Are any of the values nan or inf? Is it an image of a man holding a tench?

6.3 Memory leaks

Apparently we use normal python garbage collector analysis.

A snippet that shows all the currently allocated Tensors:

import torch
import gc
for obj in gc.get_objects():
    try:
        if torch.is_tensor(obj) or (hasattr(obj, 'data') and torch.is_tensor(obj.data)):
            print(type(obj), obj.size())
    except Exception as e:
        pass

See also usual python debugging. NB vs code has integrated PyTorch debugging support.

NB also — if any ipython line or jupyter cell returns a giant tensor it hangs around in memory, which is a very popular way to create memory leaks

6.4 Is the GPU working?

>>> import torch

>>> torch.cuda.is_available()
True

>>> torch.cuda.device_count()
1

>>> torch.cuda.current_device()
0

>>> torch.cuda.get_device_name(0)
'GeForce GTX 950M'

6.5 Logging and profiling

Leveraging TensorFlow’s handy diagnostic GUI, tensorboard: Now native, via torch.utils.tensorboard. See also the PyTorch Profiler documentation.

Easier: just use lighting if that fits the workflow.

Also I have seen visdom promoted? This pumps graphs to a visualisation server. Not PyTorch-specific, but seems well-integrated.

Further generic profiling and logging at the NN-in-practice notebook.

6.6 Visualising network graphs

Fiddly. The official way is via ONNX.

conda install -c ezyang onnx pydot # or
pip install onnx pydot

Then one can use various graphical model diagrams things.

brew install --cask netron # or
pip install netron
brew install graphviz

Also available, pytorchviz and tensorboardX support visualising PyTorch graphs.

pip install git+https://github.com/szagoruyko/pytorchviz

from pytorchviz import make_dot
y = model(x)
make_dot(y, params = dict(model.named_parameters()))

7.1 Einstein convention

Einstein convention is supported by PyTorch as torch.einsum.

Einops (Rogozhnikov 2022) is more general. It is not specific to PyTorch, but the best tutorials are for PyTorch:

• Writing better code with PyTorch+einops
• einops/2-einops-for-deep-learning

Note that there was a hyped project, Tensor Comprehensions in PyTorch (see the launch announcement) which apparently compiled the operations to CUDA kernels. It seems to be discontinued.

7.2 LinearOperator

cornellius-gp/linear_operator

A linear operator is a generalization of a matrix. It is a linear function that is defined in by its application to a vector. The most common linear operators are (potentially structured) matrices, where the function applying them to a vector are (potentially efficient) matrix-vector multiplication routines.

LinearOperator objects share (mostly) the same API as torch.Tensor objects. Under the hood, these objects use __torch_function__ to dispatch all efficient linear algebra operations to the torch and torch.linalg namespaces. […]

Each of these functions will either return a torch.Tensor, or a new LinearOperator object, depending on the function.

7.3 KeOps

The KeOps library lets you compute reductions of large arrays whose entries are given by a mathematical formula or a neural network. It combines efficient C++ routines with an automatic differentiation engine and can be used with Python (NumPy, PyTorch), Matlab and R.

It is perfectly suited to the computation of kernel matrix-vector products, K-nearest neighbors queries, N-body interactions, point cloud convolutions and the associated gradients. Crucially, it performs well even when the corresponding kernel or distance matrices do not fit into the RAM or GPU memory. Compared with a PyTorch GPU baseline, KeOps provides a x10-x100 speed-up on a wide range of geometric applications, from kernel methods to geometric deep learning.

• Kernel Operations on the GPU, with autodiff, without memory overflows
• Geometric Loss functions between sampled measures, images and volumes
• Why using KeOps?
• KernelSolve

9.1 Stochastic gradients

There is some stochastic gradient infrastructure in pyro, in the sense of differentiation through integrals, both classic score methods, reparameterisations and probably others. See, e.g. Storchastic (van Krieken, Tomczak, and Teije 2021).

• HEmile/storchastic: Stochastic Automatic Differentiation library for PyTorch.
• What is Storchastic?

9.2 ASD(FGHJK)L

kazukiosawa/asdfghjkl: ASDL: Automatic Second-order Differentiation (for Fisher, Gradient covariance, Hessian, Jacobian, and Kernel) Library (Osawa et al. 2023)

The library is called ASDL, which stands for Automatic Second-order Differentiation (for Fisher, Gradient covariance, Hessian, Jacobian, and Kernel) Library. ASDL is a PyTorch extension for computing 1st/2nd-order metrics and performing 2nd-order optimization of deep neural networks.

Used in Daxberger et al. (2021).

9.3 backpack

backpack.pt/ (Dangel, Kunstner, and Hennig 2019)

Provided quantities include:

• Individual gradients from a mini-batch
• Estimates of the gradient variance or second moment
• Approximate second-order information (diagonal and Kronecker approximations)

Motivation: Computation of most quantities is not necessarily expensive (often just a small modification of the existing backward pass where backpropagated information can be reused). But it is difficult to do in the current software environment.

Documentation mentions the following capabilities: estimate of the Variance, the Gauss-Newton Diagonal, the Gauss-Newton KFAC

Source: f-dangel/backpack.

9.4 PyHessian

amirgholami/PyHessian: PyHessian is a Pytorch library for second-order based analysis and training of Neural Networks (Yao et al. 2020):

PyHessian is a pytorch library for Hessian based analysis of neural network models. The library enables computing the following metrics:

• Top Hessian eigenvalues
• The trace of the Hessian matrix
• The full Hessian Eigenvalues Spectral Density (ESD)

9.5 Regularization gradients

One can hack the backward gradient to impose regularising penalties, but why not just use one of the pre-rolled ones by Szymon Maszke?

• szymonmaszke/torchlayers/regularization.py

10.1 Parameterizations

Lezcano/geotorch: Constrained optimization toolkit for PyTorch (Lezcano Casado 2019).

10.2 Constrained optimisation

Cooper

Cooper is a toolkit for Lagrangian-based constrained optimization in Pytorch. This library aims to encourage and facilitate the study of constrained optimization problems in machine learning.

15.1 Lightning

Lightning is a common training/utility framework for Pytorch.

Lightning is a very lightweight wrapper on PyTorch that decouples the science code from the engineering code. It’s more of a style-guide than a framework. By refactoring your code, we can automate most of the non-research code.

To use Lightning, simply refactor your research code into the LightningModule format (the science) and Lightning will automate the rest (the engineering). Lightning guarantees tested, correct, modern best practices for the automated parts.

• If you are a researcher, Lightning is infinitely flexible, you can modify everything down to the way .backward is called or distributed is set up.
• If you are a scientist or production team, lightning is very simple to use with best practice defaults.

Why do I want to use lightning?

Every research project starts the same, a model, a training loop, validation loop, etc. As your research advances, you’re likely to need distributed training, 16-bit precision, checkpointing, gradient accumulation, etc.

Lightning sets up all the boilerplate state-of-the-art training for you so you can focus on the research.

These last two paragraphs constitute a good introduction to the strengths and weaknesses of lightning: “Every research project starts the same, a model, a training loop, validation loop” stands in opposition to “Lightning is infinitely flexible”. An alternative description with different emphasis which woudl IMO be better: “Lighting can handle many ML projects that naturally factor into a single training loop but does not help so much for other projects.”

If my project does have such a factorisation, Lightning is extremely useful and will do all kinds of easy parallelisation, natural code organisation and so forth. But if I am doing something like posterior sampling, or nested iterations, or optimisation at inference time, I find myself spending more time fighting the framework than working with it.

If I want the generic scaling up, I might find myself trying one of the generic solutions like Horovod.

C&C ignite?

15.2 Fabric

• Welcome to ⚡ Fabric — lightning

Fabric differentiates itself from a fully-fledged trainer like Lightning’s Trainer in these key aspects:

Fast to implement There is no need to restructure your code: Just change a few lines in the PyTorch script and you’ll be able to leverage Fabric features.

Maximum Flexibility Write your own training and/or inference logic down to the individual optimizer calls. You aren’t forced to conform to a standardized epoch-based training loop like the one in Lightning Trainer. You can do flexible iteration based training, meta-learning, cross-validation and other types of optimization algorithms without digging into framework internals. This also makes it super easy to adopt Fabric in existing PyTorch projects to speed-up and scale your models without the compromise on large refactors. Just remember: With great power comes a great responsibility.

Maximum Control The Lightning Trainer has many built-in features to make research simpler with less boilerplate, but debugging it requires some familiarity with the framework internals. In Fabric, everything is opt-in. Think of it as a toolbox: You take out the tools (Fabric functions) you need and leave the other ones behind. This makes it easier to develop and debug your PyTorch code as you gradually add more features to it. Fabric provides important tools to remove undesired boilerplate code (distributed, hardware, checkpoints, logging,…), but leaves the design and orchestration fully up to you.

Sebastian Rascka’s blog post motivates this well: Optimising Memory Usage for Training LLMs and Vision Transformers in PyTorch.

15.3 Catalyst

I think Catalyst fills a similar niche to lightning? Not sure, have not used. The Catalyst homepage blurb seems to hit the same notes as lightning with a couple of sweeteners - e.g. it claims to support jax and tensorflow.

See also source and blogposts such as this one.

15.4 fast.ai

fastai is a deep learning library which provides practitioners with high-level components that can quickly and easily provide state-of-the-art results in standard deep learning domains, and provides researchers with low-level components that can be mixed and matched to build new approaches. It aims to do both things without substantial compromises in ease of use, flexibility, or performance. This is possible thanks to a carefully layered architecture, which expresses common underlying patterns of many deep learning and data processing techniques in terms of decoupled abstractions. These abstractions can be expressed concisely and clearly by leveraging the dynamism of the underlying Python language and the flexibility of the PyTorch library. fastai includes:

• A new type dispatch system for Python along with a semantic type hierarchy for tensors
• A GPU-optimized computer vision library which can be extended in pure Python
• An optimizer which refactors out the common functionality of modern optimizers into two basic pieces, allowing optimization algorithms to be implemented in 4–5 lines of code
• A novel 2-way callback system that can access any part of the data, model, or optimizer and change it at any point during training
• A new data block API