Skip to content

Latest commit

 

History

History
85 lines (62 loc) · 5.53 KB

README.md

File metadata and controls

85 lines (62 loc) · 5.53 KB
Raw

Grad Samples

Computing per sample gradients is an integral part of Opacus framework. We strive to provide out-of-the-box support for wide range of models, while keeping computations efficient.

We currently provide two independent approaches for computing per sample gradients: hooks-based GradSampleModule (stable implementation, exists since the very first version of Opacus) and GradSampleModuleExpandedWeights (based on a beta functionality available in PyTorch 1.12).

Each of the two implementations comes with it's own set of limitations, and we leave the choice up to the client which one to use.

GradSampleModuleExpandedWeights is currently in early beta and can produce unexpected errors, but potentially improves upon GradSampleModule on performance and functionality.

TL;DR: If you want stable implementation, use GradSampleModule (grad_sample_mode="hooks"). If you want to experiment with the new functionality, you have two options. Try GradSampleModuleExpandedWeights(grad_sample_mode="ew") for better performance and grad_sample_mode=functorch if your model is not supported by GradSampleModule.

Please switch back to GradSampleModule(grad_sample_mode="hooks") if you encounter strange errors or unexpexted behaviour. We'd also appreciate it if you report these to us

Hooks-based approach

  • Model wrapping class: opacus.grad_sample.grad_sample_module.GradSampleModule
  • Keyword argument for PrivacyEngine.make_private(): grad_sample_mode="hooks"

Computes per-sample gradients for a model using backward hooks. It requires custom grad sampler methods for every trainable layer in the model. We provide such methods for most popular PyTorch layers. Additionally, client can provide their own grad sampler for any new unsupported layer (see tutorial)

Functorch approach

  • Model wrapping class: opacus.grad_sample.grad_sample_module.GradSampleModule (force_functorch=True)
  • Keyword argument for PrivacyEngine.make_private(): grad_sample_mode="functorch"

functorch is JAX-like composable function transforms for PyTorch. With functorch we can compute per-sample-gradients efficiently by using function transforms. With the efficient parallelization provided by vmap, we can obtain per-sample gradients for any function function (i.e. any model) by doing essentially vmap(grad(f(x))).

Our experiments show, that vmap computations in most cases are as fast as manually written grad samplers used in hooks-based approach.

With the current implementation GradSampleModule will use manual grad samplers for known modules (i.e. maintain the old behaviour for all previously supported models) and will only use functorch for unknown modules.

With force_functorch=True passed to the constructor GradSampleModule will rely exclusively on functorch.

ExpandedWeigths approach

  • Model wrapping class: opacus.grad_sample.gsm_exp_weights.GradSampleModuleExpandedWeights
  • Keyword argument for PrivacyEngine.make_private(): grad_sample_mode="ew"

Computes per-sample gradients for a model using core functionality available in PyTorch 1.12+. Unlike hooks-based grad sampler, which works on a module level, ExpandedWeights work on the function level, i.e. if your layer is not explicitly supported, but only uses known operations, ExpandedWeights will support it out of the box.

At the time of writing, the coverage for custom grad samplers between GradSampleModule and GradSampleModuleExpandedWeights is roughly the same.

Comparative analysis

Please note that these are known limitations and we plan to improve Expanded Weights and bridge the gap in feature completeness

xxx Hooks Expanded Weights Functorch
Required PyTorch version 1.8+ 1.13+ 1.12 (to be updated)
Development status Underlying mechanism deprecated Beta Beta
Runtime Performance† baseline ✅ ~25% faster 🟨 0-50% slower
Any DP-allowed†† layers Not supported Not supported ✅ Supported
Most popular nn.* layers ✅ Supported ✅ Supported ✅ Supported
torchscripted models Not supported ✅ Supported Not supported
Client-provided grad sampler ✅ Supported Not supported ✅ Not needed
batch_first=False ✅ Supported Not supported ✅ Supported
Recurrent networks ✅ Supported Not supported ✅ Supported
Padding same in Conv ✅ Supported Not supported ✅ Supported
Empty poisson batches ✅ Supported Not supported Not supported

† Note, that performance differences are unstable and can vary a lot depending on the exact model and batch size. Numbers above are averaged over benchmarks with small models consisting of convolutional and linear layers. Note, that performance differences are only observed on GPU training, CPU performance seem to be almost identical for all approaches.

†† Layers that produce joint computations on batch samples (e.g. BatchNorm) are not allowed under any approach