error when compile warp_by_flow

[wxz1996]

when i compile warp_by_flow int tf1.15 with cuda10.0, it remind me the error

include "tensorflow/core/util/cuda_kernel_helper.h" not exist

it seems that tf1.15 had remove it, did you test it in any other tf version

[sylqiu]

Unfortunately I haven’t tried it with other versions of the TF.

On 10 Jul 2020, at 11:34, wxz1996 notifications@github.com wrote:

 when i compile warp_by_flow int tf1.15, it remind me the error ‘’‘#include "tensorflow/core/util/cuda_kernel_helper.h"’‘’ not exist it seems that tf1.15 had remove it, did you test it in any other tf version

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[wxz1996]

@sylqiu thanks, i have another question, i download the dataset ToFFlyingThings3D.zip and unzip it, but there has no folder "nogt" and "real" ... should i need to use the script in "Tof_pbrt_generation.zip" to generate "nogt" and "real" ?

[sylqiu]

They are loaders for loading data without ground truth for testing for instance on your own data. You can check the implementation in the data loader file.

On 10 Jul 2020, at 18:02, wxz1996 notifications@github.com wrote:

 @sylqiu thanks, i have another question, i download the dataset ToFFlyingThings3D.zip and unzip it, but there has no folder "nogt" and "real" ... should i need to use the script in "Tof_pbrt_generation.zip" to generate "nogt" and "real" ?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[wxz1996]

@sylqiu got it, but i can't find the real data, 'RealData/vivo_data/test_vivo5/' and 'RealData/vivo_data/test_vivo5/' int your rep or google driver

[sylqiu]

Unfortunately, the real dataset cannot be released, however we do have a sample included in the drive.

On 11 Jul 2020, at 17:08, wxz1996 notifications@github.com wrote:

 @sylqiu got it, but i can't find the real data, 'RealData/vivo_data/test_vivo5/' and 'RealData/vivo_data/test_vivo5/' int your rep or google driver

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[wxz1996]

@sylqiu but when i run " model.single_test()" and "dataset_name = nogt", "self.path = '/data/wenxuanzheng/ToFFlyingThings3D/' set in TestingConfig, it remind me loss real data, and i didn't find any sample end with '_reg_ir.png' and "_reg_depth.raw", is "nogt" real data sample?

[wxz1996]

···parser.add_argument('--dataset_name', type=str, default='simtof', help='dataset')
parser.add_argument('--dataset2_name', type=str, default='none', help='dataset')
parser.add_argument('--is_training', type=int, default=0, help='True:1 or False:0')
parser.add_argument('--is_from_scratch', type=int, default=0, help='Pretraining ?')
parser.add_argument('--pretrain_ckpt_path', type=str, default='none', help='path to weight file')
parser.add_argument('--save', type=str, default='./Experiment/', help='where to save stuff')
parser.add_argument('--sigma', type=str, default=0, help='perturbation strength with mvg_aug in testing mode, set 0 to use uniform perturbation, see loaders.py for definition')
parser.add_argument('--mvg_aug', type=int, default=1, help='set 1 to use multiview geometry augmentation')···

and ··· class TestingConfig(object): def init(self): self.wlast = '/home/wenxuanzheng/pc_compression/tof_rgbd_processing/Experiment/checkpoint/ckpt' self.output_save_dir ='./results' self.bs = 1

dataset path

    self.path = '/data/wenxuanzheng/ToFFlyingThings3D/'
    self.use_fifo = False

··· this is my default set, but it report cuda error"Check failed: cusolverDnCreate(&cusolver_dn_handle) == CUSOLVER_STATUS_SUCCESS Failed to create cuSolverDN instance.", and i don't know it is the compile error in custom tf_op or the parameter set error

[sylqiu]

Hmm, this is strange and I never seen it before. You can try replacing the cuda op by implementing python code counterparts.

On 13 Jul 2020, at 16:41, wxz1996 notifications@github.com wrote:

 ···parser.add_argument('--dataset_name', type=str, default='simtof', help='dataset') parser.add_argument('--dataset2_name', type=str, default='none', help='dataset') parser.add_argument('--is_training', type=int, default=0, help='True:1 or False:0') parser.add_argument('--is_from_scratch', type=int, default=0, help='Pretraining ?') parser.add_argument('--pretrain_ckpt_path', type=str, default='none', help='path to weight file') parser.add_argument('--save', type=str, default='./Experiment/', help='where to save stuff') parser.add_argument('--sigma', type=str, default=0, help='perturbation strength with mvg_aug in testing mode, set 0 to use uniform perturbation, see loaders.py for definition') parser.add_argument('--mvg_aug', type=int, default=1, help='set 1 to use multiview geometry augmentation')··· ···class TestingConfig(object): def init(self):

self.wlast = '/home/wenxuanzheng/pc_compression/tof_rgbd_processing/Experiment/checkpoint/ckpt'
self.output_save_dir ='./results'
self.bs = 1```

#dataset path
self.path = '/data/wenxuanzheng/ToFFlyingThings3D/'
self.use_fifo = False

··· this is my default set, but it report cuda error"Check failed: cusolverDnCreate(&cusolver_dn_handle) == CUSOLVER_STATUS_SUCCESS Failed to create cuSolverDN instance.", and i don't know it is the compile error in custom tf_op or the parameter set error

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[wxz1996]

nToF is the folder of real data sample? but it is not end with _reg_depth.raw ...