6.9 KiB

Raw Blame History

Model Complexity Analysis

We provide a tool to help with the complexity analysis for the network. We borrow the idea from the implementation of fvcore to build this tool, and plan to support more custom operators in the future. Currently, it provides the interfaces to compute "FLOPs", "Activations" and "Parameters", of the given model, and supports printing the related information layer-by-layer in terms of network structure or table. The analysis tool provides both operator-level and module-level flop counts simultaneously. Please refer to Flop Count for implementation details of how to accurately measure the flops of one operator if interested.

Definition

The model complexity has three indicators, namely floating-point operations (FLOPs), activations, and parameters. Their definitions are as follows:

FLOPs

Floating-point operations (FLOPs) is not a clearly defined indicator. Here, we refer to the description in detectron2, which defines a set of multiply-accumulate operations as 1 FLOP.
Activations

Activation is used to measure the feature quantity produced from one layer.
Parameters

The parameter count of a model.

For example, given an input size of inputs = torch.randn((1, 3, 10, 10)) and a convolutional layer conv = nn.Conv2d(in_channels=3, out_channels=10, kernel_size=3), if the output feature map size is (1, 10, 8, 8), then its FLOPs are 17280 = 10*8*8*3*3*3 (where 10*8*8 represents the output feature map size, and 3*3*3 represents the computation for each output), activations are 640 = 10*8*8, and the parameter count is 280 = 3*10*3*3 + 10 (where 3*10*3*3 represents the size of weights, and 10 represents the size of bias).

Usage

Model built with native nn.Module

Build a model

from torch import nn
from mmengine.analysis import get_model_complexity_info


# return a dict of analysis results, including:
# ['flops', 'flops_str', 'activations', 'activations_str', 'params', 'params_str', 'out_table', 'out_arch']

class InnerNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc1 = nn.Linear(10,10)
        self.fc2 = nn.Linear(10,10)
    def forward(self, x):
        return self.fc1(self.fc2(x))


class TestNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc1 = nn.Linear(10,10)
        self.fc2 = nn.Linear(10,10)
        self.inner = InnerNet()
    def forward(self, x):
        return self.fc1(self.fc2(self.inner(x)))

input_shape = (1, 10)
model = TestNet()

analysis_results = get_model_complexity_info(model, input_shape)

The analysis_results returned by get_model_complexity_info is a dict, which contains the following keys:

flops: number of total flops, e.g., 10000, 10000
flops_str: with formatted string, e.g., 1.0G, 100M
params: number of total parameters, e.g., 10000, 10000
params_str: with formatted string, e.g., 1.0G, 100M
activations: number of total activations, e.g., 10000, 10000
activations_str: with formatted string, e.g., 1.0G, 100M
out_table: print related information by table

Print the results

print related information by table

print(analysis_results['out_table'])

+---------------------+----------------------+--------+--------------+
| module              | #parameters or shape | #flops | #activations |
+---------------------+----------------------+--------+--------------+
| model               | 0.44K                | 0.4K   | 40           |
|  fc1                |  0.11K               |  100   |  10          |
|   fc1.weight        |   (10, 10)           |        |              |
|   fc1.bias          |   (10,)              |        |              |
|  fc2                |  0.11K               |  100   |  10          |
|   fc2.weight        |   (10, 10)           |        |              |
|   fc2.bias          |   (10,)              |        |              |
|  inner              |  0.22K               |  0.2K  |  20          |
|   inner.fc1         |   0.11K              |   100  |   10         |
|    inner.fc1.weight |    (10, 10)          |        |              |
|    inner.fc1.bias   |    (10,)             |        |              |
|   inner.fc2         |   0.11K              |   100  |   10         |
|    inner.fc2.weight |    (10, 10)          |        |              |
|    inner.fc2.bias   |    (10,)             |        |              |
+---------------------+----------------------+--------+--------------+

print related information by network layers

print(analysis_results['out_arch'])

TestNet(
  #params: 0.44K, #flops: 0.4K, #acts: 40
  (fc1): Linear(
    in_features=10, out_features=10, bias=True
    #params: 0.11K, #flops: 100, #acts: 10
  )
  (fc2): Linear(
    in_features=10, out_features=10, bias=True
    #params: 0.11K, #flops: 100, #acts: 10
  )
  (inner): InnerNet(
    #params: 0.22K, #flops: 0.2K, #acts: 20
    (fc1): Linear(
      in_features=10, out_features=10, bias=True
      #params: 0.11K, #flops: 100, #acts: 10
    )
    (fc2): Linear(
      in_features=10, out_features=10, bias=True
      #params: 0.11K, #flops: 100, #acts: 10
    )
  )
)

print results with formatted string

print("Model Flops:{}".format(analysis_results['flops_str']))
# Model Flops:0.4K
print("Model Parameters:{}".format(analysis_results['params_str']))
# Model Parameters:0.44K

Model built with mmengine

import torch.nn.functional as F
import torchvision
from mmengine.model import BaseModel
from mmengine.analysis import get_model_complexity_info


class MMResNet50(BaseModel):
    def __init__(self):
        super().__init__()
        self.resnet = torchvision.models.resnet50()

    def forward(self, imgs, labels=None, mode='tensor'):
        x = self.resnet(imgs)
        if mode == 'loss':
            return {'loss': F.cross_entropy(x, labels)}
        elif mode == 'predict':
            return x, labels
        elif mode == 'tensor':
            return x


input_shape = (3, 224, 224)
model = MMResNet50()

analysis_results = get_model_complexity_info(model, input_shape)

print("Model Flops:{}".format(analysis_results['flops_str']))
# Model Flops:4.145G
print("Model Parameters:{}".format(analysis_results['params_str']))
# Model Parameters:25.557M

Interface

We provide more options to support custom output

model: (nn.Module) the model to be analyzed
input_shape: (tuple) the shape of the input, e.g., (3, 224, 224)
inputs: (optional: torch.Tensor), if given, input_shape will be ignored
show_table: (bool) whether return the statistics in the form of table, default: True
show_arch: (bool) whether return the statistics by network layers, default: True

6.9 KiB Raw Blame History