This article will familiarize you with the transformation process of Python - > Cafe - > om model

Hua Weiyun 2021-02-23 02:22:36
article familiarize transformation process python

Abstract : This article aims to share Pytorch->Caffe->om Model transformation process .

Standard network


The main function code is :

+-- Caffe
| +-- caffe.proto
| +--
+-- example
| +--

Direct use can refer to, If the operation in the network Baseline Zhongdu has been realized , Can be converted directly to Caffe Model .

Add custom actions

If you encounter an operation that is not implemented , There are two situations to consider .

Caffe There are corresponding operations in

With arg_max For example, let's share how to add operations .

The first thing to look at is Caffe The parameters of the corresponding layer in :caffe.proto For the corresponding version caffe The definition of layers and parameters , You can see ArgMax Defined out_max_val、top_k、axis Three parameters :

message ArgMaxParameter {
// If true produce pairs (argmax, maxval)
optional bool out_max_val = 1 [default = false];
optional uint32 top_k = 2 [default = 1];
// The axis along which to maximise -- may be negative to index from the
// end (e.g., -1 for the last axis).
// By default ArgMaxLayer maximizes over the flattened trailing dimensions
// for each index of the first / num dimension.
optional int32 axis = 3;

And Caffe Operator boundary The parameters in are consistent . Build the instance of parameter class in concrete transformation , Realized the operation parameter from Pytorch To Caffe The transfer :

def argmax_param(self, out_max_val=None, top_k=None, dim=1):
argmax_param = pb.ArgMaxParameter()
if out_max_val is not None:
argmax_param.out_max_val = out_max_val
if top_k is not None:
argmax_param.top_k = top_k
if dim is not None:
argmax_param.axis = dim
self.param.argmax_param.CopyFrom(argmax_param) It defines Rp class , Used to implement Pytorch Operate to Caffe The transformation of operation :

class Rp(object):
def __init__(self, raw, replace, **kwargs):
self.obj = replace
self.raw = raw
def __call__(self, *args, **kwargs):
return self.raw(*args, **kwargs)
for stack in traceback.walk_stack(None):
if 'self' in stack[0].f_locals:
layer = stack[0].f_locals['self']
if layer in layer_names:
log.pytorch_layer_name = layer_names[layer]
print('984', layer_names[layer])
out = self.obj(self.raw, *args, **kwargs)
return out

When adding operations , To use Rp Class replacement operation :

torch.argmax = Rp(torch.argmax, torch_argmax)

Next , To implement this operation :

def torch_argmax(raw, input, dim=1):
x = raw(input, dim=dim)
layer_name = log.add_layer(name='argmax')
top_blobs = log.add_blobs([x], name='argmax_blob'.format(type))
layer = caffe_net.Layer_param(name=layer_name, type='ArgMax',
bottom=[log.blobs(input)], top=top_blobs)
return x

It is realized. argmax operation Pytorch To Caffe Transformation .

Caffe There is no direct corresponding operation in

If the operation to be converted is in Caffe There is no directly corresponding layer implementation in , There are two main solutions :

1) stay Pytorch Decompose unsupported operations into supported operations :

Such as nn.InstanceNorm2d, Instance normalization is used in conversion BatchNorm It's done , I won't support it affine=True perhaps track_running_stats=True, Default use_global_stats:false, but om On conversion use_global_stats It has to be for true, So you can go to Caffe, But turn around om unfriendly .

InstanceNorm Is in featuremap Each Channel Normalization operation is carried out on , therefore , Can achieve nn.InstanceNorm2d by :

class InstanceNormalization(nn.Module):
def __init__(self, dim, eps=1e-5):
super(InstanceNormalization, self).__init__()
self.gamma = nn.Parameter(torch.FloatTensor(dim))
self.beta = nn.Parameter(torch.FloatTensor(dim))
self.eps = eps
def _reset_parameters(self):
def __call__(self, x):
n = x.size(2) * x.size(3)
t = x.view(x.size(0), x.size(1), n)
mean = torch.mean(t, 2).unsqueeze(2).unsqueeze(3).expand_as(x)
var = torch.var(t, 2).unsqueeze(2).unsqueeze(3).expand_as(x)
gamma_broadcast = self.gamma.unsqueeze(1).unsqueeze(1).unsqueeze(0).expand_as(x)
beta_broadcast = self.beta.unsqueeze(1).unsqueeze(1).unsqueeze(0).expand_as(x)
out = (x - mean) / torch.sqrt(var + self.eps)
out = out * gamma_broadcast + beta_broadcast
return out

But in verification HiLens Caffe Operator boundary Found in ,om Model transformation does not support Channle Sum or mean operations outside of dimensions , To circumvent this operation , We can reimplement it with supported operators nn.InstanceNorm2d:

class InstanceNormalization(nn.Module):
def __init__(self, dim, eps=1e-5):
super(InstanceNormalization, self).__init__()
self.gamma = torch.FloatTensor(dim)
self.beta = torch.FloatTensor(dim)
self.eps = eps
self.adavg = nn.AdaptiveAvgPool2d(1)
def forward(self, x):
n, c, h, w = x.shape
mean = nn.Upsample(scale_factor=h)(self.adavg(x))
var = nn.Upsample(scale_factor=h)(self.adavg((x - mean).pow(2)))
gamma_broadcast = self.gamma.unsqueeze(1).unsqueeze(1).unsqueeze(0).expand_as(x)
beta_broadcast = self.beta.unsqueeze(1).unsqueeze(1).unsqueeze(0).expand_as(x)
out = (x - mean) / torch.sqrt(var + self.eps)
out = out * gamma_broadcast + beta_broadcast
return out

After verification , Equivalent to the original operation , It can be changed to Caffe Model

2) stay Caffe By using existing operations to achieve :

stay Pytorch turn Caffe In the process of discovery , If there is featuremap + 6 This involves constant operations , In the process of conversion, it will appear that blob The problem of . Let's first look at in add Specific conversion method of operation :

def _add(input, *args):
x = raw__add__(input, *args)
return x
layer_name = log.add_layer(name='add')
top_blobs = log.add_blobs([x], name='add_blob')
if log.blobs(args[0]) == None:
log.add_blobs([args[0]], name='extra_blob')
layer = caffe_net.Layer_param(name=layer_name, type='Eltwise',
bottom=[log.blobs(input),log.blobs(args[0])], top=top_blobs)
layer.param.eltwise_param.operation = 1 # sum is 1
return x

You can see that for blob There is no case to judge , All we need to do is log.blobs(args[0]) == None To modify under certain conditions , A natural idea is to use Scale Layer implementation add operation :

def _add(input, *args):
x = raw__add__(input, *args)
return x
layer_name = log.add_layer(name='add')
top_blobs = log.add_blobs([x], name='add_blob')
if log.blobs(args[0]) == None:
layer = caffe_net.Layer_param(name=layer_name, type='Scale',
bottom=[log.blobs(input)], top=top_blobs)
layer.param.scale_param.bias_term = True
weight = torch.ones((input.shape[1]))
bias = torch.tensor(args[0]).squeeze().expand_as(weight)
layer.add_data(weight.cpu().data.numpy(), bias.cpu().data.numpy())
layer = caffe_net.Layer_param(name=layer_name, type='Eltwise',
bottom=[log.blobs(input), log.blobs(args[0])], top=top_blobs)
layer.param.eltwise_param.operation = 1 # sum is 1
return x

Allied ,featuremap * 6 This simple multiplication can be done in the same way .

The pit of tread

  • Pooling:Pytorch Default ceil_mode=false,Caffe Default ceil_mode=true, It can lead to dimensional changes , If there is a size mismatch, you can check Pooling Is the parameter correct . in addition , Although not seen in the document , however kernel_size > 32 Although the post model can be transformed , But reasoning will report errors , This can be done in two layers Pooling operation .
  • Upsample :om In the boundary operator Upsample layer scale_factor Parameter must be int, It can't be size. If the existing model parameter is size It's going to run normally Pytorch turn Caffe The process of , But this time Upsample Parameter is empty . Parameter is size It can be considered that scale_factor Or use Deconvolution To achieve .
  • Transpose2d:Pytorch in output_padding Parameters are added to the size of the output , but Caffe Can't , The output feature map will be smaller , Now, after deconvolution featuremap It's going to get bigger , Can pass Crop The layers are cut , Make it the same size as Pytorch The corresponding layers are consistent . in addition ,om The speed of deconvolution reasoning is slow , It's better not to use , It can be used Upsample+Convolution replace .
  • Pad:Pytorch in Pad There are many different operations , but Caffe Can only be carried out in H And W Symmetry in dimensions pad, If Pytorch There is h = F.pad(x, (1, 2, 1, 2), "constant", 0) This asymmetry pad operation , The solution is as follows :
  1. If it's asymmetric pad There is no subsequent dimension mismatch problem in the layer of , You can judge first pad The effect on the result , Some tasks are affected by pad The impact is very small , Then there's no need to modify .
  2. If there is a dimension mismatch problem , It can be considered that according to the larger parameters pad after Crop, Or the front and back (0, 0, 1, 1) And (1, 1, 0, 0) Of pad In one (1, 1, 1, 1), It depends on the specific network structure .
  3. If it is Channel Dimensionally pad Such as F.pad(x, (0, 0, 0, 0, 0, channel_pad), "constant", 0), We can consider zero convolution cat To featuremap On :
zero = nn.Conv2d(in_channels, self.channel_pad, kernel_size=3, padding=1, bias=False)
nn.init.constant(, 0)
pad_tensor = zero(x)
x =[x, pad_tensor], dim=1)
  • Some operations can go to Caffe, but om Standards are not supported Caffe All operations , If you want to go to om Check the boundary operator against the document .
This article is shared from Huawei cloud community 《Pytorch->Caffe Model transformation 》, Original author : Du Fu built a house .


Click to follow , The first time to learn about Huawei's new cloud technology ~

本文为[Hua Weiyun]所创,转载请带上原文链接,感谢

  1. 利用Python爬虫获取招聘网站职位信息
  2. Using Python crawler to obtain job information of recruitment website
  3. Several highly rated Python libraries arrow, jsonpath, psutil and tenacity are recommended
  4. Python装饰器
  5. Python实现LDAP认证
  6. Python decorator
  7. Implementing LDAP authentication with Python
  8. Vscode configures Python development environment!
  9. In Python, how dare you say you can't log module? ️
  10. 我收藏的有关Python的电子书和资料
  11. python 中 lambda的一些tips
  12. python中字典的一些tips
  13. python 用生成器生成斐波那契数列
  14. python脚本转pyc踩了个坑。。。
  15. My collection of e-books and materials about Python
  16. Some tips of lambda in Python
  17. Some tips of dictionary in Python
  18. Using Python generator to generate Fibonacci sequence
  19. The conversion of Python script to PyC stepped on a pit...
  20. Python游戏开发,pygame模块,Python实现扫雷小游戏
  21. Python game development, pyGame module, python implementation of minesweeping games
  22. Python实用工具,email模块,Python实现邮件远程控制自己电脑
  23. Python utility, email module, python realizes mail remote control of its own computer
  24. 毫无头绪的自学Python,你可能连门槛都摸不到!【最佳学习路线】
  25. Python读取二进制文件代码方法解析
  26. Python字典的实现原理
  27. Without a clue, you may not even touch the threshold【 Best learning route]
  28. Parsing method of Python reading binary file code
  29. Implementation principle of Python dictionary
  30. You must know the function of pandas to parse JSON data - JSON_ normalize()
  31. Python实用案例,私人定制,Python自动化生成爱豆专属2021日历
  32. Python practical case, private customization, python automatic generation of Adu exclusive 2021 calendar
  33. 《Python实例》震惊了,用Python这么简单实现了聊天系统的脏话,广告检测
  34. "Python instance" was shocked and realized the dirty words and advertisement detection of the chat system in Python
  35. Convolutional neural network processing sequence for Python deep learning
  36. Python data structure and algorithm (1) -- enum type enum
  37. 超全大厂算法岗百问百答(推荐系统/机器学习/深度学习/C++/Spark/python)
  38. 【Python进阶】你真的明白NumPy中的ndarray吗?
  39. All questions and answers for algorithm posts of super large factories (recommended system / machine learning / deep learning / C + + / spark / Python)
  40. [advanced Python] do you really understand ndarray in numpy?
  41. 【Python进阶】Python进阶专栏栏主自述:不忘初心,砥砺前行
  42. [advanced Python] Python advanced column main readme: never forget the original intention and forge ahead
  43. python垃圾回收和缓存管理
  44. java调用Python程序
  45. java调用Python程序
  46. Python常用函数有哪些?Python基础入门课程
  47. Python garbage collection and cache management
  48. Java calling Python program
  49. Java calling Python program
  50. What functions are commonly used in Python? Introduction to Python Basics
  51. Python basic knowledge
  52. Anaconda5.2 安装 Python 库(MySQLdb)的方法
  53. Python实现对脑电数据情绪分析
  54. Anaconda 5.2 method of installing Python Library (mysqldb)
  55. Python implements emotion analysis of EEG data
  56. Master some advanced usage of Python in 30 seconds, which makes others envy it
  57. python爬取百度图片并对图片做一系列处理
  58. Python crawls Baidu pictures and does a series of processing on them
  59. python链接mysql数据库
  60. Python link MySQL database