Releases: 1adrianb/pytorch-estimate-flops
v0.4.1 released
[Add] Added support for FX based estimation for pytorch 1.8 and newer, jit based approach is deprecated.
[Fix] Fixed 2 issues with the jit based estimator (@senysenyseny16)
v0.3.5 released
[Fix] Added support for pytorch 1.8.0
v0.3.4 released
[Fix] Issue when the input is a list
[Fix] Regex problem when the tensor was on a cuda device
v0.3.0 released
[Fix] Added support for pytorch 1.6.0
[Fix] Fix ignore flag issues with pytorch 1.4.0 and newer
[Fix] Custom ops were previously ignored
v0.2.2 released
[Fix] Fixed compatibility with pytorch 1.4.0
v0.2.1 released
[Change] A detailed dictionary containing each op is now also returned
[Add] Added support for mul, global average pooling and matmul
v0.2.0 released
[Fix] Issue with ndConvs (thanks Jean!)
[Add] Added option to display the flops for each individual layer
[Add] Added option to ignore certain module types
[Add] Added option to provide custom functions for computing the OPs
Initial release
Initial release with the main structure and a few basic ops