torch-weighttracker 0.1.4

Tools for tracking structured weight sparsity in PyTorch models.

torch-weighttracker

Tools for tracking structured weight sparsity, regularization signals, and bit-operation estimates in PyTorch models.

The package builds a structural view of a model, compiles tensorized reduction plans over that structure, and reuses those plans for training-time metrics and regularizers.

import torch
from torch import nn

from torch_weighttracker import WeightTracker

model = nn.Sequential(nn.Linear(4, 8), nn.ReLU(), nn.Linear(8, 2))
tracker = WeightTracker(model, example_inputs=torch.randn(1, 4))

print(tracker.view_structures())

TRANSFORMS NOT FULLY SUPPORTED YET

Installation

python -m pip install torch-weighttracker

Structured BOPs MAC accounting uses fvcore for baseline per-module MACs:

python -m pip install "torch-weighttracker[structured-bops]"

Why Use It?

PyTorch makes it easy to inspect one parameter tensor at a time. Structured compression often needs a different view:

• A channel can be coupled across convolutions, batch norms, linear layers, and residual paths.
• A transformer unit can mean an attention head, a head dimension, or a fused QKV slice rather than a simple row or column.
• A metric such as "active BOPs" depends on sparsity, module shape, MAC counts, and bitrates at the same time.
• A regularizer such as group lasso should penalize the coupled structural unit, not each weight tensor independently.

WeightTracker turns those coupled structures into canonical units, then lets calculations operate over the canonical units with reusable tensor programs.

Use Cases

Current use cases:

• Add structured group lasso to a training loss.
• Track active structured BOPs and compression rate during structured pruning, sparsity-aware training, or quantization-aware training (QAT).
• Inspect which modules participate in each channel, feature, head, or head-dim group.
• Build structural metrics that aggregate many weight tensors into one value per pruning unit.

Group Lasso

Structured group lasso regularizes coupled units together. Layers can be excluded per regularizer:

from torch_weighttracker.regularizers import RegularizerType

group_lasso = tracker.create_regularizer(
    RegularizerType.GROUP_LASSO,
    ignore=[model.classifier],
)

loss = task_loss + 1e-4 * group_lasso()
loss.backward()

Structured BOPs

Structured BOPs reports compression against a dense 32-bit baseline by default:

import torch

from torch_weighttracker.trackers import TrackerType

metrics = tracker.create_tracker(
    TrackerType.STRUCTURED_BOPS,
    include=[model.layer3, model.layer4],
    ignore=[torch.nn.BatchNorm2d],
).track()

print(metrics["structured_bops_compression"])
print(metrics["structured_bops_compression_rate_pr_module"])

raw_metrics = tracker.create_tracker(
    TrackerType.STRUCTURED_BOPS,
    include=[model.layer3, model.layer4],
    ignore=[torch.nn.BatchNorm2d],
    log_total_bops=True,
).track()

print(raw_metrics["structured_bops"])
print(raw_metrics["structured_bops_pr_module"])

create_tracker accepts a single TrackerType/string or a list of tracker types/strings:

tracker.create_tracker([TrackerType.STRUCTURED_BOPS, "unstructured_sparsity"])
metrics = tracker.track()

Unstructured Sparsity

Unstructured sparsity reports exact zero-weight fractions. The total is weighted by each layer's number of weight elements, not averaged across layer fractions:

import torch

from torch_weighttracker.trackers import TrackerType

metrics = tracker.create_tracker(
    TrackerType.UNSTRUCTURED_SPARSITY,
    include=[model.layer3, model.layer4],
    ignore=[torch.nn.BatchNorm2d],
).track()

print(metrics["unstructured_sparsity"])
print(metrics["layers"])

Values are fractions in [0, 1]. Parametrized fake quantization is measured through the effective module.weight, so quantized zeros count as sparse weights.

Formal defintion

Structured BOPs are calculated from the pruned layerwise MACs and the bitwidths of the weights and activations [1]:

[ \text{BOPs}l = \text{MACs}l \cdot b{w,l} \cdot b{a,l-1} ]

The pruned layerwise MACs are:

[ \text{MACs}l = (1 - p{l-1})c_{l-1} \cdot (1 - p_l)c_l \cdot m_{w,l} \cdot m_{h,l} \cdot k_w \cdot k_h ]

The effective layerwise pruning ratio is:

[ P_l = 1 - (1 - p_{l-1})(1 - p_l) ]

Where:

• (\text{BOPs}_l): bit operations for layer (l)
• (\text{MACs}_l): multiply-accumulate operations for layer (l)
• (b_{w,l}): weight bitwidth for layer (l)
• (b_{a,l-1}): activation bitwidth from the previous layer
• (p_{l-1}): pruning ratio of the input structures
• (p_l): pruning ratio of the output structures
• (P_l): effective pruning ratio for layer (l)
• (c_{l-1}): number of input channels
• (c_l): number of output channels
• (m_{w,l}, m_{h,l}): output feature map width and height
• (k_w, k_h): kernel width and height

Comparison with Direct Removal and FLOP Count

For some model architectures, the BOPs calculation may differ from the values reported by other libraries. These differences are mainly due to variations in which layers and operations are included in the calculation. WeightTracker does not account for elementwise operations such as ReLU activations or bias terms.

As a sanity check, I have included notebooks that illustrate these differences by comparing fvcore.FlopCountAnalysis on hard-pruned models with WeightTracker on fake-pruned models, where weights are zeroed such that it is equivalent to the hard-pruned.

Local sanity notebooks compare WeightTracker MAC accounting with physically pruned models from Torch-Pruning. These dependencies are optional and are not installed with the base package:

python -m pip install -e ".[dev-local]"

Then start Jupyter from the repository root and open the notebooks in sanity_checks/.

Architecture

The main API is WeightTracker. Internally it is split into a few layers:

1. Dependency discovery: WeightTracker builds dependency groups from the model and example_inputs, or accepts precomputed groups.
2. Canonical units: canonical_units.py normalizes raw dependency groups into CanonicalUnitGroup objects. These give channels, features, attention heads, and head dimensions a shared unit index.
3. Reduction plans: reductions/ and plans/ compile module and unit mappings into segment and index operations that use PyTorch's efficient tensor computations.
4. Calculations: calculations/ defines named calculation specs such as per-unit L2 norm, active units, parameters per unit, active MACs, and bitrates. Calculations can depend on each other and cache constant results.
5. Consumers: regularizers/ and trackers/ request the calculations they need, optionally with include and ignore contexts for selecting modules in a specific metric or regularizer.

The result is a small public surface with a reusable internal graph:

model + example inputs
        |
        v
dependency groups -> canonical units -> reduction plans -> calculations
                                                          |
                                                          v
                                               regularizers and trackers

Speed

Compared with a naive implementation, the current implementation gives the following speedups on ResNet 20 on a RTX 3060:

Comparison	Speedup	Naive extra allocation	WeightTracker extra allocation
Group lasso	15.421x	197.0MiB	197.0MiB
Structured BOPs	2.582x	1.7GiB	195.9MiB

Status

This package is pre-1.0. Public APIs may still change while the tracker, calculation, and regularizer surfaces settle.

Future Work

1. Streamline definitions and method names across the codebase.
2. Improve calculation caching so repeated computations are not performed twice.
3. Improve compilation of computation plans for bigger speedups.
4. Improve memory management within calculations.
5. Write more comprehensive docstrings.

Future custom use cases will need a broader top-level WeightTracker API for custom operations, custom layers, and generic group definitions.

License

MIT

References

[1] Wang et al., Differentiable Joint Pruning and Quantization for Hardware Efficiency, 2020.