Skip to content

Fix ConvTranspose #150

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Aug 20, 2025
Merged

Fix ConvTranspose #150

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4e9de26
Fix ConvTranspoe
FrzMtrsprt Aug 20, 2025
aba9753
Merge branch 'master' into fix-convt
sovrasov Aug 20, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 12 additions & 6 deletions ptflops/pytorch_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,8 @@
* this file. If not visit https://opensource.org/licenses/MIT
'''

from functools import partial

import numpy as np
import torch
import torch.nn as nn
Expand Down Expand Up @@ -55,12 +57,11 @@ def bn_flops_counter_hook(module, input, output):
module.__flops__ += int(batch_flops)


def conv_flops_counter_hook(conv_module, input, output, extra_per_position_flops=0):
def conv_flops_counter_hook(conv_module, input, output, extra_per_position_flops=0, transpose=False):
# Can have multiple inputs, getting the first one
input = input[0]

batch_size = input.shape[0]
output_dims = list(output.shape[2:])

kernel_dims = list(conv_module.kernel_size)
in_channels = conv_module.in_channels
Expand All @@ -71,7 +72,12 @@ def conv_flops_counter_hook(conv_module, input, output, extra_per_position_flops
conv_per_position_flops = int(np.prod(kernel_dims, dtype=np.int64)) * \
(in_channels * filters_per_channel + extra_per_position_flops)

active_elements_count = batch_size * int(np.prod(output_dims, dtype=np.int64))
if transpose:
input_dims = list(input.shape[2:])
active_elements_count = batch_size * int(np.prod(input_dims, dtype=np.int64))
else:
output_dims = list(output.shape[2:])
active_elements_count = batch_size * int(np.prod(output_dims, dtype=np.int64))

overall_conv_flops = conv_per_position_flops * active_elements_count

Expand Down Expand Up @@ -301,9 +307,9 @@ def timm_attention_counter_hook(attention_module, input, output):
# Upscale
nn.Upsample: upsample_flops_counter_hook,
# Deconvolution
nn.ConvTranspose1d: conv_flops_counter_hook,
nn.ConvTranspose2d: conv_flops_counter_hook,
nn.ConvTranspose3d: conv_flops_counter_hook,
nn.ConvTranspose1d: partial(conv_flops_counter_hook, transpose=True),
nn.ConvTranspose2d: partial(conv_flops_counter_hook, transpose=True),
nn.ConvTranspose3d: partial(conv_flops_counter_hook, transpose=True),
# RNN
nn.RNN: rnn_flops_counter_hook,
nn.GRU: rnn_flops_counter_hook,
Expand Down
Loading