pytorch version #44

WangYibucea · 2020-06-11T18:31:23Z

Do you have pytorch version?

hazdzz · 2020-11-10T20:23:08Z

The author does not have a PyTorch version of STGCN. However, there are 4 implementations made by PyTorch by others. Unfortunately, all those PyTorch implementations do not follow the paper.

jeongwhanchoi · 2020-11-11T05:37:00Z

The author does not have a PyTorch version of STGCN. However, there are 4 implementations made by PyTorch by others. Unfortunately, all those PyTorch implementations do not follow the paper.

Would you know those 4 implementations are different from the original STGCN?

hazdzz · 2020-11-13T19:58:02Z

The author does not have a PyTorch version of STGCN. However, there are 4 implementations made by PyTorch by others. Unfortunately, all those PyTorch implementations do not follow the paper.

Would you know those 4 implementations are different from the original STGCN?

hazdzz · 2020-11-24T14:31:22Z

I have made a PyTorch version. Here is my code: https://github.com/hazdzz/STGCN

wanzhixiao · 2020-11-27T09:14:33Z

我做了一个PyTorch版本，没有做任何修改。这是我的代码：https : //github.com/hazdzz/STGCN

how do you guarantee your implementation follow the paper?

wanzhixiao · 2020-11-27T10:50:01Z

does this implementation of temporal convoution module correct?

class temporal_conv_layer(nn.Module):

def __init__(self, Kt, c_in, c_out):
    '''
    :param Kt: int, kernel size of temporal convolution.
    :param c_in: int, size of input channel.
    :param c_out: int, size of output channel.
    '''
    super(temporal_conv_layer,self).__init__()
    self.Kt = Kt
    self.c_in = c_in
    self.c_out = c_out

    self.conv1 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(1,1),stride=1,padding=0)
    self.conv2 = nn.Conv2d(in_channels=c_in,out_channels=2*c_out,kernel_size=(Kt,1),stride=1)
    self.conv3 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(Kt,1),stride=1)

    self.act = F.relu

def forward(self,x,act_func):
    '''
    :param x: input tensor ,[batch_size, c_in,time_step, n_route]
    :return: tensor, [batch_size, c_out,time_step-Kt+1, n_route].
    '''
    _,c,T,N = x.shape
    if(self.c_in > self.c_out): 
        x_input = self.conv1(x)
    elif self.c_in < self.c_out:
        # if the size of input channel is less than the output,
        # padding x to the same size of output channel.
        # Note, _.get_shape() cannot convert a partially known TensorShape to a Tensor.
        x_input = torch.cat(x,torch.zeros((x.shape[0],self.c_out-c,T,N)),dim=1)
    else:
        x_input = x
    # keep the original input for residual connection.
    x_input = x_input[:,:,self.Kt - 1:T, :]

    if act_func == 'GLU':
        # gated liner unit
        # GLU h(x) = (X*W+b)×sigmod(X*V+c)
        x_conv = self.conv2(x)
        #residual *sigmod(x)
        #(x+F(x))*sigmod(x)
        return (x_conv[:,0:self.c_out,:,:] + x_input) * torch.sigmoid(x_conv[:,-self.c_out:,:,:])
    else:

        x_conv = self.conv3(x)
        if act_func == 'linear':
            return x_conv
        elif act_func == 'sigmoid':
            return torch.sigmoid(x_conv)
        elif act_func == 'relu':
            return self.act(x_conv + x_input)
        else:
            raise ValueError(f'ERROR: activation function "{act_func}" is not defined.')

hazdzz · 2020-11-27T13:44:22Z

我做了一个PyTorch版本，没有做任何修改。这是我的代码：https : //github.com/hazdzz/STGCN

how do you guarantee your implementation follow the paper?

I translated the code from the author's which used TensorFlow v1 to mine which used PyTorch. That's it. And I found the author misunderstands the Laplacian matrix in GCN(first-order approximation) which I fixed.

hazdzz · 2020-11-27T14:19:53Z

does this implementation of temporal convoution module correct?

class temporal_conv_layer(nn.Module):

def __init__(self, Kt, c_in, c_out):
    '''
    :param Kt: int, kernel size of temporal convolution.
    :param c_in: int, size of input channel.
    :param c_out: int, size of output channel.
    '''
    super(temporal_conv_layer,self).__init__()
    self.Kt = Kt
    self.c_in = c_in
    self.c_out = c_out

    self.conv1 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(1,1),stride=1,padding=0)
    self.conv2 = nn.Conv2d(in_channels=c_in,out_channels=2*c_out,kernel_size=(Kt,1),stride=1)
    self.conv3 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(Kt,1),stride=1)

    self.act = F.relu

def forward(self,x,act_func):
    '''
    :param x: input tensor ,[batch_size, c_in,time_step, n_route]
    :return: tensor, [batch_size, c_out,time_step-Kt+1, n_route].
    '''
    _,c,T,N = x.shape
    if(self.c_in > self.c_out): 
        x_input = self.conv1(x)
    elif self.c_in < self.c_out:
        # if the size of input channel is less than the output,
        # padding x to the same size of output channel.
        # Note, _.get_shape() cannot convert a partially known TensorShape to a Tensor.
        x_input = torch.cat(x,torch.zeros((x.shape[0],self.c_out-c,T,N)),dim=1)
    else:
        x_input = x
    # keep the original input for residual connection.
    x_input = x_input[:,:,self.Kt - 1:T, :]

    if act_func == 'GLU':
        # gated liner unit
        # GLU h(x) = (X*W+b)×sigmod(X*V+c)
        x_conv = self.conv2(x)
        #residual *sigmod(x)
        #(x+F(x))*sigmod(x)
        return (x_conv[:,0:self.c_out,:,:] + x_input) * torch.sigmoid(x_conv[:,-self.c_out:,:,:])
    else:

        x_conv = self.conv3(x)
        if act_func == 'linear':
            return x_conv
        elif act_func == 'sigmoid':
            return torch.sigmoid(x_conv)
        elif act_func == 'relu':
            return self.act(x_conv + x_input)
        else:
            raise ValueError(f'ERROR: activation function "{act_func}" is not defined.')

Yes, it is correct.

wanzhixiao · 2020-11-28T02:39:36Z

does this implementation of temporal convoution module correct?
class temporal_conv_layer(nn.Module):

def __init__(self, Kt, c_in, c_out):
    '''
    :param Kt: int, kernel size of temporal convolution.
    :param c_in: int, size of input channel.
    :param c_out: int, size of output channel.
    '''
    super(temporal_conv_layer,self).__init__()
    self.Kt = Kt
    self.c_in = c_in
    self.c_out = c_out

    self.conv1 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(1,1),stride=1,padding=0)
    self.conv2 = nn.Conv2d(in_channels=c_in,out_channels=2*c_out,kernel_size=(Kt,1),stride=1)
    self.conv3 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(Kt,1),stride=1)

    self.act = F.relu

def forward(self,x,act_func):
    '''
    :param x: input tensor ,[batch_size, c_in,time_step, n_route]
    :return: tensor, [batch_size, c_out,time_step-Kt+1, n_route].
    '''
    _,c,T,N = x.shape
    if(self.c_in > self.c_out): 
        x_input = self.conv1(x)
    elif self.c_in < self.c_out:
        # if the size of input channel is less than the output,
        # padding x to the same size of output channel.
        # Note, _.get_shape() cannot convert a partially known TensorShape to a Tensor.
        x_input = torch.cat(x,torch.zeros((x.shape[0],self.c_out-c,T,N)),dim=1)
    else:
        x_input = x
    # keep the original input for residual connection.
    x_input = x_input[:,:,self.Kt - 1:T, :]

    if act_func == 'GLU':
        # gated liner unit
        # GLU h(x) = (X*W+b)×sigmod(X*V+c)
        x_conv = self.conv2(x)
        #residual *sigmod(x)
        #(x+F(x))*sigmod(x)
        return (x_conv[:,0:self.c_out,:,:] + x_input) * torch.sigmoid(x_conv[:,-self.c_out:,:,:])
    else:

        x_conv = self.conv3(x)
        if act_func == 'linear':
            return x_conv
        elif act_func == 'sigmoid':
            return torch.sigmoid(x_conv)
        elif act_func == 'relu':
            return self.act(x_conv + x_input)
        else:
            raise ValueError(f'ERROR: activation function "{act_func}" is not defined.')

Yes, it is correct.

Why do we do one-dimensional convolution in feature dimension instead of time dimension

if(self.c_ in > self.c_ out):

x_ input = self.conv1 (x)

hazdzz · 2020-11-28T03:01:13Z

does this implementation of temporal convoution module correct?

class temporal_conv_layer(nn.Module):

def init(self, Kt, c_in, c_out):

'''

:param Kt: int, kernel size of temporal convolution.

:param c_in: int, size of input channel.

:param c_out: int, size of output channel.

'''

super(temporal_conv_layer,self).__init__()

self.Kt = Kt

self.c_in = c_in

self.c_out = c_out

self.conv1 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(1,1),stride=1,padding=0)

self.conv2 = nn.Conv2d(in_channels=c_in,out_channels=2*c_out,kernel_size=(Kt,1),stride=1)

self.conv3 = nn.Conv2d(in_channels=c_in,out_channels=c_out,kernel_size=(Kt,1),stride=1)

self.act = F.relu

def forward(self,x,act_func):

'''

:param x: input tensor ,[batch_size, c_in,time_step, n_route]

:return: tensor, [batch_size, c_out,time_step-Kt+1, n_route].

'''

_,c,T,N = x.shape

if(self.c_in > self.c_out):

    x_input = self.conv1(x)

elif self.c_in < self.c_out:

    # if the size of input channel is less than the output,

    # padding x to the same size of output channel.

    # Note, _.get_shape() cannot convert a partially known TensorShape to a Tensor.

    x_input = torch.cat(x,torch.zeros((x.shape[0],self.c_out-c,T,N)),dim=1)

else:

    x_input = x

# keep the original input for residual connection.

x_input = x_input[:,:,self.Kt - 1:T, :]

if act_func == 'GLU':

    # gated liner unit

    # GLU h(x) = (X*W+b)×sigmod(X*V+c)

    x_conv = self.conv2(x)

    #residual *sigmod(x)

    #(x+F(x))*sigmod(x)

    return (x_conv[:,0:self.c_out,:,:] + x_input) * torch.sigmoid(x_conv[:,-self.c_out:,:,:])

else:

    x_conv = self.conv3(x)

    if act_func == 'linear':

        return x_conv

    elif act_func == 'sigmoid':

        return torch.sigmoid(x_conv)

    elif act_func == 'relu':

        return self.act(x_conv + x_input)

    else:

        raise ValueError(f'ERROR: activation function "{act_func}" is not defined.')

Yes, it is correct.

Why do we do one-dimensional convolution in feature dimension instead of time dimension

if(self.c_ in > self.c_ out):

x_ input = self.conv1 (x)

I think you should ask the author. I have no idea either.

qiubin1997 · 2021-03-10T03:47:13Z

作者没有STGCN的PyTorch版本。但是，其他人通过PyTorch进行了4种实现。不幸的是，所有这些PyTorch的实现都没有遵循本文。

请问一下我用的windows10运行的程序中的main.py 总是报错UnboundLocalError: local variable 'W' referenced before assignment（赋值前引用的局部变量W）的问题，如何修改，请指教一下。谢谢。

D:\Anaconda\envs\tensorflow-gpu\python.exe D:/PycharmProjects/STGCN_IJCAI-18/STGCN_IJCAI-18-master/main.py
2021-03-10 11:39:36.031711: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
2021-03-10 11:39:37.332511: I tensorflow/core/platform/cpu_feature_guard.cc:143] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2021-03-10 11:39:37.338870: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x287b22ae980 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2021-03-10 11:39:37.339049: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
2021-03-10 11:39:37.339710: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library nvcuda.dll
2021-03-10 11:39:37.362410: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1561] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: GeForce GTX 1050 Ti computeCapability: 6.1
coreClock: 1.392GHz coreCount: 6 deviceMemorySize: 4.00GiB deviceMemoryBandwidth: 104.43GiB/s
2021-03-10 11:39:37.362663: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
2021-03-10 11:39:37.365917: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_10.dll
2021-03-10 11:39:37.368303: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_10.dll
2021-03-10 11:39:37.369087: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_10.dll
2021-03-10 11:39:37.372096: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_10.dll
2021-03-10 11:39:37.373803: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_10.dll
2021-03-10 11:39:37.381498: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2021-03-10 11:39:37.381689: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1703] Adding visible gpu devices: 0
2021-03-10 11:39:37.845596: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-03-10 11:39:37.845754: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1108] 0
2021-03-10 11:39:37.845836: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1121] 0: N
2021-03-10 11:39:37.846049: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1247] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 2988 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1050 Ti, pci bus id: 0000:01:00.0, compute capability: 6.1)
2021-03-10 11:39:37.848793: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x287d618f8c0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2021-03-10 11:39:37.848965: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): GeForce GTX 1050 Ti, Compute Capability 6.1
Training configs: Namespace(batch_size=50, epoch=50, graph='default', inf_mode='merge', ks=3, kt=3, lr=0.001, n_his=12, n_pred=9, n_route=228, opt='RMSProp', save=10)
ERROR: input file was not found in ./dataset\PeMSD7_W_228.csv.
Traceback (most recent call last):
File "D:/PycharmProjects/STGCN_IJCAI-18/STGCN_IJCAI-18-master/main.py", line 52, in
W = weight_matrix(pjoin('./dataset', f'PeMSD7_W_{n}.csv'))
File "D:\PycharmProjects\STGCN_IJCAI-18\STGCN_IJCAI-18-master\utils\math_graph.py", line 87, in weight_matrix
if set(np.unique(W)) == {0, 1}:
UnboundLocalError: local variable 'W' referenced before assignment

Process finished with exit code 1

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

pytorch version #44

pytorch version #44

WangYibucea commented Jun 11, 2020

hazdzz commented Nov 10, 2020

jeongwhanchoi commented Nov 11, 2020

hazdzz commented Nov 13, 2020

hazdzz commented Nov 24, 2020 •

edited

Loading

wanzhixiao commented Nov 27, 2020 •

edited

Loading

wanzhixiao commented Nov 27, 2020 •

edited

Loading

hazdzz commented Nov 27, 2020 •

edited

Loading

hazdzz commented Nov 27, 2020

wanzhixiao commented Nov 28, 2020

hazdzz commented Nov 28, 2020

qiubin1997 commented Mar 10, 2021

pytorch version #44

pytorch version #44

Comments

WangYibucea commented Jun 11, 2020

hazdzz commented Nov 10, 2020

jeongwhanchoi commented Nov 11, 2020

hazdzz commented Nov 13, 2020

hazdzz commented Nov 24, 2020 • edited Loading

wanzhixiao commented Nov 27, 2020 • edited Loading

wanzhixiao commented Nov 27, 2020 • edited Loading

hazdzz commented Nov 27, 2020 • edited Loading

hazdzz commented Nov 27, 2020

wanzhixiao commented Nov 28, 2020

hazdzz commented Nov 28, 2020

qiubin1997 commented Mar 10, 2021

hazdzz commented Nov 24, 2020 •

edited

Loading

wanzhixiao commented Nov 27, 2020 •

edited

Loading

wanzhixiao commented Nov 27, 2020 •

edited

Loading

hazdzz commented Nov 27, 2020 •

edited

Loading