Skip to content

Latest commit

 

History

History
87 lines (64 loc) · 2.31 KB

CHANGELOG.md

File metadata and controls

87 lines (64 loc) · 2.31 KB

ptflops versions log

v 0.7.4

  • Fix hook for nn.functional.interpolate.
  • Add ignore and custom modules for aten.
  • Add an option to disable counting of functional-style operations in pytorch backend.

v 0.7.3

  • Add aten backend to collect the amount of flops on aten level.

v 0.7.2.2

  • Switch from setup.py to pyproject

v 0.7.2

  • Add type annotations and doc strings to the main API.
  • Add support of HuggingFace/Timm VIT transformers.
  • Update torchvision benchmark in docs.

v 0.7.1.2

  • Fix failure when using input constructor.

v 0.7.1

  • Experimental support of torchvision.ops.DeformConv2d
  • Experimental support of torch.functional.* and tensor.* operators

v 0.7

  • Add ConvNext to sample, fix wrong torchvision compatibility requirement.
  • Support LayerNorm.

v 0.6.9

  • Fix unnecessary warnings.
  • Improve per layer statistics output.

v 0.6.8

  • Add support of GELU activation.
  • Fix per layer statistic output in case of zero parameters number.
  • Cleanup flops and params attrs after ptflops has finished counting.

v 0.6.7

  • Add batch_first flag support in MultiheadAttention hook

v 0.6.6

  • Add hooks for Instance and Group norms.

v 0.6.5

  • Add a hook for MultiheadAttention.

v 0.6.4

  • Fix unaccounted bias flops in Linear.
  • Fix hook for ConvTranspose*d.

v 0.6.3

  • Implicitly use repr to print a model with extra_repr.

v 0.6.2

  • Fix integer overflow on Windows.
  • Check if the input object is inherited from nn.Module.

v 0.6.1

  • Add experimental version of hooks for recurrent layers (RNN, GRU, LSTM).

v 0.6

  • Add verbose option to log layers that are not supported by ptflops.
  • Add an option to filter a list of operations from the final result.

v 0.5.2

  • Fix handling of intermediate dimensions in the Linear layer hook.

v 0.5

  • Add per sequential number of parameters estimation.
  • Fix sample doesn't work without GPU.
  • Clarified output in sample.

v 0.4

  • Allocate temporal blobs on the same device as model's parameters are located.

v 0.3

  • Add 1d operators: batch norm, poolings, convolution.
  • Add ability to output extended report to any output stream.

v 0.2

  • Add new operations: Conv3d, BatchNorm3d, MaxPool3d, AvgPool3d, ConvTranspose2d.
  • Add some results on widespread models to the README.
  • Minor bugfixes.

v 0.1

  • Initial release with basic functionality