[*] Reading checkpoints... [!] Load failed... about dcgan-tensorflow HOT 27 CLOSED

Hi!

I just ran into this, and I found the problem.

What happens, is that if you just leave the out_name flag as default, it will just create a new folder, and then it will try to load the checkpoint from there.

So the solution is to set the out_name flag to a previous model folder. Either from the code or as a command line parameter. Like
--out_name="20190818.161208 - data - test - x64.z100.uniform_signed.y64.b64"

Peace

from dcgan-tensorflow.

Hi

I am also facing this loading fail issue. Can anyone help me with this.

I successfully trained model but fails at testing.

[] **Reading checkpoints... ./out/20190712.155740 - data - face - x96.z100.uniform_signed.y96.b64/checkpoint
[] Failed to find a checkpoint
[!] Load failed...**
[ 1 Epoch:[ 0/300] [ 0/ 1] time: 7.3325, d_loss: 12.05198956, g_loss: 0.00001405
[ 2 Epoch:[ 1/300] [ 0/ 1] time: 12.2742, d_loss: 10.16373634, g_loss: 0.00009138
[ 3 Epoch:[ 2/300] [ 0/ 1] time: 17.1198, d_loss: 12.54014683, g_loss: 0.00000996
[ 4 Epoch:[ 3/300] [ 0/ 1] time: 22.0227, d_loss: 12.31657314, g_loss: 0.00003112
[ 5 Epoch:[ 4/300] [ 0/ 1] time: 26.9523, d_loss: 12.80654812, g_loss: 0.00008534
[ 6 Epoch:[ 5/300] [ 0/ 1] time: 31.8358, d_loss: 13.03531933, g_loss: 0.00001151
[ 7 Epoch:[ 6/300] [ 0/ 1] time: 36.8203, d_loss: 13.17454529, g_loss: 0.00003603
[ 8 Epoch:[ 7/300] [ 0/ 1] time: 41.7034, d_loss: 13.25488567, g_loss: 0.00006080
[ 9 Epoch:[ 8/300] [ 0/ 1] time: 46.6821, d_loss: 12.10309505, g_loss: 0.00041392
[ 10 Epoch:[ 9/300] [ 0/ 1] time: 51.5538, d_loss: 9.41581345, g_loss: 0.00186340
[ 11 Epoch:[10/300] [ 0/ 1] time: 56.5072, d_loss: 10.10222530, g_loss: 0.00050865
[ 12 Epoch:[11/300] [ 0/ 1] time: 61.3804, d_loss: 4.94413757, g_loss: 0.14236367
[ 13 Epoch:[12/300] [ 0/ 1] time: 66.4153, d_loss: 11.87597275, g_loss: 0.00002049
[ 14 Epoch:[13/300] [ 0/ 1] time: 71.3269, d_loss: 1.26149607, g_loss: 3.08130121
[ 15 Epoch:[14/300] [ 0/ 1] time: 76.1823, d_loss: 12.14586639, g_loss: 0.00001769
[ 16 Epoch:[15/300] [ 0/ 1] time: 81.1462, d_loss: 1.42883611, g_loss: 1.98843646
[ 17 Epoch:[16/300] [ 0/ 1] time: 85.9693, d_loss: 11.67339706, g_loss: 0.00007065
[ 18 Epoch:[17/300] [ 0/ 1] time: 90.9558, d_loss: 0.86891878, g_loss: 5.06913090
[ 19 Epoch:[18/300] [ 0/ 1] time: 96.0543, d_loss: 10.42274475, g_loss: 0.00019286
[ 20 Epoch:[19/300] [ 0/ 1] time: 101.0552, d_loss: 1.43598580, g_loss: 8.86531544
[ 21 Epoch:[20/300] [ 0/ 1] time: 106.2137, d_loss: 3.21956253, g_loss: 0.13176931
[ 22 Epoch:[21/300] [ 0/ 1] time: 111.1346, d_loss: 4.63180637, g_loss: 0.03671605
[ 23 Epoch:[22/300] [ 0/ 1] time: 115.9590, d_loss: 0.81582117, g_loss: 11.40536022
[ 24 Epoch:[23/300] [ 0/ 1] time: 120.7342, d_loss: 1.90553343, g_loss: 0.64395195
[ 25 Epoch:[24/300] [ 0/ 1] time: 125.7073, d_loss: 8.43229103, g_loss: 0.00032412
[ 26 Epoch:[25/300] [ 0/ 1] time: 130.6088, d_loss: 0.95951736, g_loss:

Training is fine but testing is not being done. throwing error of :

raise Exception("Checkpoint not found in " + FLAGS.checkpoint_dir)
Exception: Checkpoint not found in ./out/20190712.103742 - data - face/checkpoint

I have trained on my own dataset.

from dcgan-tensorflow.

@carpedm20 Thanks for the code. It is raising crazy errors:
TypeError: DataType float32 for attr 'Tshape' not in list of allowed values: int32, int64 at
self.z_, [-1, s_h16, s_w16, self.gf_dim * 8]) /model.py in genarator
Can you share your implementation

from dcgan-tensorflow.

If there is no checkpoint of a model, that's a normal output not an error.

from dcgan-tensorflow.

@carpedm20 Thank you very much ! Do you have a email??

from dcgan-tensorflow.

i got the same error,as below:

[*] Reading checkpoints...
 [*] Failed to find a checkpoint
 [!] Load failed...
 [*] 0
 [*] 1
 [*] 2
 [*] 3
 [*] 4
 [*] 5
 [*] 6
 [*] 7
 [*] 8
 [*] 9
 [*] 10
 [*] 11
 [*] 12
 [*] 13
 [*] 14
 [*] 15
 [*] 16
 [*] 17
 [*] 18
 [*] 19
 [*] 20
 [*] 21
 [*] 22
 [*] 23
 [*] 24
 [*] 25
 [*] 26
 [*] 27
 [*] 28
 [*] 29
 [*] 30
 [*] 31
 [*] 32
 [*] 33
 [*] 34
 [*] 35
 [*] 36
 [*] 37
 [*] 38
 [*] 39
 [*] 40
 [*] 41
 [*] 42
 [*] 43
 [*] 44
 [*] 45
 [*] 46
 [*] 47
 [*] 48
 [*] 49
 [*] 50
 [*] 51
 [*] 52
 [*] 53
 [*] 54
 [*] 55
 [*] 56
 [*] 57
 [*] 58
 [*] 59
 [*] 60
 [*] 61
 [*] 62
 [*] 63
 [*] 64
 [*] 65
 [*] 66
 [*] 67
 [*] 68
 [*] 69
 [*] 70
 [*] 71
 [*] 72
 [*] 73
 [*] 74
 [*] 75
 [*] 76
 [*] 77
 [*] 78
 [*] 79
 [*] 80
 [*] 81
 [*] 82
 [*] 83
 [*] 84
 [*] 85
 [*] 86
 [*] 87
 [*] 88
 [*] 89
 [*] 90
 [*] 91
 [*] 92
 [*] 93
 [*] 94
 [*] 95
 [*] 96
 [*] 97
 [*] 98
 [*] 99

from dcgan-tensorflow.

@shartoo Did you trained the model first? Aren't you trying to test (generate samples) a model without any training?

from dcgan-tensorflow.

@carpedm20 no, the output above is exactly what the training step print when training celeA

from dcgan-tensorflow.

@shartoo I am getting same error can you suggest, how to proceed further. Can you share the checkpoint

from dcgan-tensorflow.

@carpedm20 this error occurs when your train data set contains nothing,you should check it out.

from dcgan-tensorflow.

@carpedm20 I have given path to celebA dataset and it is able to take date. It is failing to get a checkpoint. Can you suggest how to train and get our own checkpoint.

from dcgan-tensorflow.

@carpedm20 the code should be ran in two stages, training and generate. The training stage will create checkpoint directory and generate stage will load it. When running the code at first time,you should change it for training .In main.py

flags.DEFINE_boolean("is_train", False, "True for training, False for testing [False]")

If the variable is_train is True,it's training step and generate step if it was False.

flags.DEFINE_string("dataset", "celebA", "The name of dataset [celebA, mnist, lsun]")

the celebA to the your dataset name below current directory ./data/ and take care of the image size,they must be 64*64. My code for crop is below(some functions are copied from utils.py of the project):

# -*- coding:utf-8 -*-
'''
crop the image to  square
'''
import scipy.misc
import numpy as np
import glob
import os

def center_crop(x, crop_h, crop_w=None, resize_w=64):
    if crop_w is None:
        crop_w = crop_h
    h, w = x.shape[:2]
    j = int(round((h - crop_h)/2.))
    i = int(round((w - crop_w)/2.))
    return scipy.misc.imresize(x[j:j+crop_h, i:i+crop_w],
                               [resize_w, resize_w])

def imread(path, is_grayscale = False):
    if (is_grayscale):
        return scipy.misc.imread(path, flatten = True).astype(np.float)
    else:
        return scipy.misc.imread(path).astype(np.float)

def imsave(image, path):
    return scipy.misc.imsave(path,image)

if __name__=='__main__':
    data = glob(os.path.join("./data","human_face", "*.jpg"))
    for file in data:
        imsave(center_crop(imread(file,False),64),file)
        print("crop image %s saved.."%file)

from dcgan-tensorflow.

I am working on super resolution algorithm based on DCGAN. I am getting same error like Reading checkpoints , load failed .. I changed the code as suggested by you.

from dcgan-tensorflow.

@Chaitu1509 I got the same error， did u solve that？ it is really strange. almost crazy

from dcgan-tensorflow.

I mean that 'Tshape' one

from dcgan-tensorflow.

@Chaitu1509 I solved it ,see

#104

from dcgan-tensorflow.

requirement

from dcgan-tensorflow.

@shartoo i have already done based on your code but still the same error
$ python main.py --dataset=cloth --input_fname_pattern=".png" --c_dim=1 --is_train=True
[] Reading checkpoints...
[] Failed to find a checkpoint
[!] Load failed...
[] 0
[] 1
[] 2
[] 3
[] 4
[] 5
[] 6
[] 7
[] 8
[] 9
[] 10
[] 11
[] 12
[] 13
[] 14
[] 15
[] 16
[] 17
[] 18
[] 19
[] 20
[] 21
[] 22
[] 23
[] 24
[] 25
[] 26
[] 27
[] 28
[] 29
[] 30
[] 31
[] 32
[] 33
[] 34
[] 35
[] 36
[] 37
[] 38
[] 39
[] 40
[] 41
[] 42
[] 43
[] 44
[] 45
[] 46
[] 47
[] 48
[] 49
[] 50
[] 51
[] 52
[] 53
[] 54
[] 55
[] 56
[] 57
[] 58
[] 59
[] 60
[] 61
[] 62
[] 63
[] 64
[] 65
[] 66
[] 67
[] 68
[] 69
[] 70
[] 71
[] 72
[] 73
[] 74
[] 75
[] 76
[] 77
[] 78
[] 79
[] 80
[] 81
[] 82
[] 83
[] 84
[] 85
[] 86
[] 87
[] 88
[] 89
[] 90
[] 91
[] 92
[] 93
[] 94
[] 95
[] 96
[] 97
[] 98
[*] 99

from dcgan-tensorflow.

@tewea check the command you ran,the parameter --is_train should be --is_train=True?

from dcgan-tensorflow.

@shartoo still it is the same error even i did --is_train=True is there something I can change in the crop code? my issue is i have my own dataset i create new cloth folder under ./data and I modify flags.DEFINE_string("dataset", "cloth", "The name of dataset [cloth, mnist, lsun]") as well but still getting the same error

from dcgan-tensorflow.

@tewea this error has nothing to do with crop code, it could be caused by missing correct dataset ,maybe you could put your images to path ./data/cloth and keep same as other dataset like lsun.

from dcgan-tensorflow.

@shartoo you are right it works fine with load successful for "celebA" after I repeat twice but still i have issue when i use small pictures but still it shows only empty Grey png pictures in sample folder. how can i implement your crop code in which file name? and still not working with my own dataset I try to copy to the path ./data/cloth my images are downloaded manually and cropped 64*64 size few of them manually for test purpose from http://people.ee.ethz.ch/~lbossard/projects/accv12/index.html site

from dcgan-tensorflow.

It was a long time since my last debugging ,this repository seems not work well for new trianing dataset,i'll sends my code to your email and post my code here after some modify.Please leave your email address.

from dcgan-tensorflow.

Hi

@shartoo My email id is [email protected].. please send me your code. This repo is not working. Checkpoint is not saved after complete training on my dataset and hence Testing is not possible.

Thanks in advance Shartoo.

from dcgan-tensorflow.

I am also having this issue

from dcgan-tensorflow.

from dcgan-tensorflow.

Hi!

I just ran into this, and I found the problem.

What happens, is that if you just leave the out_name flag as default, it will just create a new folder, and then it will try to load the checkpoint from there.

So the solution is to set the out_name flag to a previous model folder. Either from the code or as a command line parameter. Like
--out_name="20190818.161208 - data - test - x64.z100.uniform_signed.y64.b64"

Peace
thank u so much!
Under ur help, i'v solved my problem!

from dcgan-tensorflow.