Skip to content

Edit input_shape to input.shape #2

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

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Edit input_shape to input.shape #2

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
951d045
Update compute_layer_flops.py so it match new tf versions
ZanderNic Oct 28, 2024
c07defd
bugfix
ZanderNic Oct 28, 2024
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
35 changes: 16 additions & 19 deletions flopco_keras/compute_layer_flops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,3 @@

def numel(w : list):
out = 1
for k in w:
Expand All @@ -20,27 +19,26 @@ def compute_conv2d_flops(layer, macs = False):

# _, cin, h, w = input_shape
if layer.data_format == "channels_first":
_, input_channels, _, _ = layer.input_shape
_, output_channels, h, w, = layer.output_shape
_, input_channels, _, _ = layer.input.shape
_, output_channels, h, w, = layer.output.shape
elif layer.data_format == "channels_last":
_, _, _, input_channels = layer.input_shape
_, h, w, output_channels = layer.output_shape
_, _, _, input_channels = layer.input.shape
_, h, w, output_channels = layer.output.shape

w_h, w_w = layer.kernel_size

# flops = h * w * output_channels * input_channels * w_h * w_w / (stride**2)
flops = h * w * output_channels * input_channels * w_h * w_w


if not macs:
flops_bias = numel(layer.output_shape[1:]) if layer.use_bias is not None else 0
flops_bias = numel(layer.output.shape[1:]) if layer.use_bias is not None else 0
flops = 2 * flops + flops_bias

return int(flops)


def compute_fc_flops(layer, macs = False):
ft_in, ft_out = layer.input_shape[-1], layer.output_shape[-1]
ft_in, ft_out = layer.input.shape[-1], layer.output.shape[-1]
flops = ft_in * ft_out

if not macs:
Expand All @@ -51,19 +49,18 @@ def compute_fc_flops(layer, macs = False):

def compute_bn2d_flops(layer, macs = False):
# subtract, divide, gamma, beta
flops = 2 * numel(layer.input_shape[1:])
flops = 2 * numel(layer.input.shape[1:])

if not macs:
flops *= 2

return int(flops)


def compute_relu_flops(layer, macs = False):

flops = 0
if not macs:
flops = numel(layer.input_shape[1:])
flops = numel(layer.input.shape[1:])

return int(flops)

Expand All @@ -72,7 +69,7 @@ def compute_maxpool2d_flops(layer, macs = False):

flops = 0
if not macs:
flops = layer.pool_size[0]**2 * numel(layer.output_shape[1:])
flops = layer.pool_size[0]**2 * numel(layer.output.shape[1:])

return flops

Expand All @@ -81,24 +78,24 @@ def compute_pool2d_flops(layer, macs = False):

flops = 0
if not macs:
flops = layer.pool_size[0]**2 * numel(layer.output_shape[1:])
flops = layer.pool_size[0]**2 * numel(layer.output.shape[1:])

return flops

def compute_globalavgpool2d_flops(layer, macs = False):

if layer.data_format == "channels_first":
_, input_channels, h, w = layer.input_shape
_, output_channels = layer.output_shape
_, input_channels, h, w = layer.input.shape
_, output_channels = layer.output.shape
elif layer.data_format == "channels_last":
_, h, w, input_channels = layer.input_shape
_, output_channels = layer.output_shape
_, h, w, input_channels = layer.input.shape
_, output_channels = layer.output.shape

return h*w

def compute_softmax_flops(layer, macs = False):

nfeatures = numel(layer.input_shape[1:])
nfeatures = numel(layer.input.shape[1:])

total_exp = nfeatures # https://stackoverflow.com/questions/3979942/what-is-the-complexity-real-cost-of-exp-in-cmath-compared-to-a-flop
total_add = nfeatures - 1
Expand Down