trypag

training with 273 patches
validating with 682 patches
architecture: ModSegNet
number of workers: 4
number of epochs: 170
starting epoch: 0
batch size: 8
learning rate: 0.001000
learning rate decay: 0.00100
momentum: 0.90000

trypag

import tensor_comprehensions as tc
import torch


def main():
    lang = """
    def graph2(float(N, C, H, W) I, float(N, C, L, M) J, float(KH, KW) W1) -> (O, G) {{
        O(n, c, h, w) +=! I(n, c, h + kh, w + kw) * W1(kh, kw)
        G(c, c) +=! I(n, c, h, w) * J(n, c, h, w)
    }}

trypag

import torch
from torch.autograd import Variable


def accuracy_2d(output, target, topk=(1,)):
    """
    Computes the precision@k for the specified values of k
    Considers output is : NxCxHxW and target is : NxHxW
    """
    maxk = max(topk)

trypag

def train(train_loader, model, criterion, optimizer, epoch, logger):
    batch_time = AverageMeter()
    data_time = AverageMeter()
    losses = AverageMeter()
    top1 = AverageMeter()
    top3 = AverageMeter()

    # switch to train mode
    model.train()

trypag

def accuracy_2d(output, target, topk=(1,)):
    """
    Computes the precision@k for the specified values of k

    Considers output is : NxCxHxW and target is : NxHxW
    """
    maxk = max(topk)
    batch_size = target.size(0)
    total_nelem = batch_size*target.size(-1)*target.size(-2)

trypag

Keybase proof

I hereby claim:

• I am trypag on github.
• I am trypag (https://keybase.io/trypag) on keybase.
• I have a public key whose fingerprint is 4538 4367 0091 3EE2 47EC 605F 85CD C328 F74C F395

To claim this, I am signing this object: