lanpa

trace = torch.jit.trace(model_alexnet, args_alexnet)
graph = trace.graph
print(graph)
============================================
.graph(%input.1 : Double(2, 3, 224, 224),
      %196 : Tensor,
      %197 : Tensor,
      %198 : Tensor,
      %199 : Tensor,
      %200 : Tensor,

lanpa

graph(%0 : Float(1, 3)
      %1 : Float(5, 3)
      %2 : Float(5)) {
  %3 : Dynamic = onnx::Constant[value={0}]()
  %4 : Dynamic = onnx::Gemm[alpha=1, beta=0, transB=1](%0, %1, %3)
  %5 : Float(1, 5) = onnx::Add(%2, %4)
  return (%5);
}

%3 : Dynamic = onnx::Constant[value={0}]()

lanpa

running install
running bdist_egg
running egg_info
writing tensorboardX.egg-info/PKG-INFO
writing dependency_links to tensorboardX.egg-info/dependency_links.txt
writing requirements to tensorboardX.egg-info/requires.txt
writing top-level names to tensorboardX.egg-info/top_level.txt
reading manifest file 'tensorboardX.egg-info/SOURCES.txt'
reading manifest template 'MANIFEST.in'
writing manifest file 'tensorboardX.egg-info/SOURCES.txt'

lanpa

running install
Done generating tensorboardX/proto/*pb2*.py
running build
running build_py
copying tensorboardX/proto/types_pb2.py -> build/lib/tensorboardX/proto
copying tensorboardX/proto/resource_handle_pb2.py -> build/lib/tensorboardX/proto
copying tensorboardX/proto/event_pb2.py -> build/lib/tensorboardX/proto
copying tensorboardX/proto/summary_pb2.py -> build/lib/tensorboardX/proto
copying tensorboardX/proto/graph_pb2.py -> build/lib/tensorboardX/proto
copying tensorboardX/proto/layout_pb2.py -> build/lib/tensorboardX/proto

lanpa

---------------------------  ---------------  ---------------  ---------------  ---------------  ---------------
Name                                CPU time        CUDA time            Calls        CPU total       CUDA total
---------------------------  ---------------  ---------------  ---------------  ---------------  ---------------
conv2d                             462.645us          0.000us                1        462.645us          0.000us
convolution                        461.316us          0.000us                1        461.316us          0.000us
_convolution                       459.809us          0.000us                1        459.809us          0.000us
tensor                               2.967us          0.000us                1          2.967us          0.000us
_convolution_nogroup               445.746us          0.000us                1        445.746us          0.000us
thnn_conv2d                        441.713us          0.000us                1        441.713us          0.000us

lanpa

import torch.nn as nn
import torch

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(6,9)
    def forward(self, x):
        return self.fc(x.view(-1, 6))

lanpa

class Net1(nn.Module):
    def __init__(self):
        super(Net1, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)
        self.bn = nn.BatchNorm2d(20)

lanpa

class Net1(nn.Module):
    def __init__(self):
        super(Net1, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)
        self.bn = nn.BatchNorm2d(20)

lanpa

from __future__ import print_function
import argparse
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.autograd import Variable
from tensorboardX import SummaryWriter
# Training settings

lanpa

with open ('data.csv') as f:
    lines = f.readlines()

valid = []
for line in lines:
    a, b = line.strip('\n').split(',')
    if a=='1' or a=='2' or a=='3':
        print(a, b)
        valid.append((int(a), int(b)))