sparseml icon indicating copy to clipboard operation
sparseml copied to clipboard

Libraries for applying sparsification recipes to neural networks with a few lines of code, enabling faster and smaller models

Results 165 sparseml issues
Sort by recently updated
recently updated
newest added

When running the imagenet training script from torchvision (i.e `src/sparseml/pytorch/image_classification/train.py`) I face the deadlock issue at certain random moment of training (in between the training epoch, usually somewhere in between...

bug

This PR represents the dev branch for all `sparseml.recipe_template` related code; This PR is in draft mode and will be merged into main on feature completion, The PR has been...

mle-team

Bumps [protobuf](https://github.com/protocolbuffers/protobuf) from 3.17.3 to 3.18.3. Release notes Sourced from protobuf's releases. Protocol Buffers v3.18.3 C++ Reduce memory consumption of MessageSet parsing This release addresses a Security Advisory for C++...

dependencies

a core feature of the QuantizationModifier refactor is the ability for users to have both more simple and more fine grained control over how quantization is applied at large and...

mle-team

This PR adds testing support for analyzer models and node shape for both torch 1.9 and 1.12. - Each has a data directory with both `torch-1.9` and `torch-1.12` directories. -...

mle-team

first PR for QuantizationModifier refactor. moves the existing modifier and tests into a "legacy" file. Creates a template object for the new modifier. To maintain backwards compatibility, we add support...

mle-team

This PR adds a `ModifierBuildInfo` class with state + utility methods for building a recipe from `modifier(s)` This also adds a registry for various pruning and quantization modifiers, that will...

mle-team

This PR adds a CLI for `sparseml.recipe_template` + automated tests TODO: - [ ] Change base to `sparsify.recipe_template.dev` before landing

mle-team

This PR adds getters for fetching `ModifierBuildInfo` objects from registry based off of args specified by user TODO: - [ ] Change base to `sparseml.recipe_template.dev` before landing

mle-team

This feature improves OBSPruningModifier by allowing to recompute inverse Hessian `num_recomputations` times when doing a pruning step. For example, in the standard pipeline we would ask OBSPruningModifier to prune a...