aurotripathy

name: "spacenet_SSD_300x300_train"
layer {
  name: "data"
  type: "AnnotatedData"
  top: "data"
  top: "label"
  include {
    phase: TRAIN
  }
  transform_param {

aurotripathy

"""
Time the all_reduce_sum operation with model parameters size of the VGG-16 model (~138M floats)
    1. Paramters are loaded into each of the N=4 GPUs
    2. nccl.all_reduce is invoked on the paramters
TO get a breakdown of the VGG model size, see...
https://stackoverflow.com/questions/28232235/how-to-calculate-the-number-of-parameters-of-convolutional-neural-networks
"""

from __future__ import print_function
import torch

aurotripathy

import json
from pudb import set_trace
import matplotlib
import matplotlib.pyplot as plt
import numpy as np


with open('good_run_raport.json') as json_file:
    data = json.load(json_file)
    print("top 5")

aurotripathy

""
This script does a quick-and-dirty validation of the ImageNet dataset.
TODO - a quick EDA (exploratory data analysis), which is customary before any analysis
"""
import os
import glob
import imghdr
import argparse

nb_classes = 1000

aurotripathy

""" Demonstrates the easy of integration of a custom layer """
import math
import torch
import torch.nn as nn
import numpy as np

class MyLinearLayer(nn.Module):
    """ Custom Linear layer but mimics a standard linear layer """
    def __init__(self, size_in, size_out):
        super().__init__()

aurotripathy

""" Demonstrates the easy of integration of a custom layer """
import math
import torch
import torch.nn as nn
import numpy as np

aurotripathy

class MyLinearLayer(nn.Module):
    """ Custom Linear layer but mimics a standard linear layer """
    def __init__(self, size_in, size_out):
        super().__init__()
        self.size_in, self.size_out = size_in, size_out
        weights = torch.Tensor(size_out, size_in)
        self.weights = nn.Parameter(weights)  # nn.Parameter is a Tensor that's a module parameter.
        bias = torch.Tensor(size_out)
        self.bias = nn.Parameter(bias)

aurotripathy

class MyLinearLayer(nn.Module):
    """ Custom Linear layer but mimics a standard linear layer """
    def __init__(self, size_in, size_out):
        super().__init__()
        self.size_in, self.size_out = size_in, size_out
        weights = torch.Tensor(size_out, size_in)
        self.weights = nn.Parameter(weights)  # nn.Parameter is a Tensor that's a module parameter.
        bias = torch.Tensor(size_out)
        self.bias = nn.Parameter(bias)

aurotripathy

class BasicModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv = nn.Conv2d(1, 128, 3)
        # self.linear = nn.Linear(256, 2)
        self.linear = MyLinearLayer(256, 2)

    def forward(self, x):
        x = self. conv(x)
        x = x.view(-1, 256)

aurotripathy

torch.manual_seed(0)  #  for repeatable results
basic_model = BasicModel()
inp = np.array([[[[1,2,3,4],  # batch(=1) x channels(=1) x height x width
                  [1,2,3,4],
                  [1,2,3,4]]]])
x = torch.tensor(inp, dtype=torch.float)
print('Forward computation thru model:', basic_model(x))