facebookresearch / SparseConvNet

Submanifold sparse convolutional networks
https://github.com/facebookresearch/SparseConvNet
Other
2.03k stars 332 forks source link

Problem with SparseConvNet/sparseconvnet/metadata.py #48

Closed csershen closed 6 years ago

csershen commented 6 years ago

SparseConvNet/sparseconvnet/metadata.py

Has a problem when running on the gpu. Should the underscore be a slash?

btgraham commented 6 years ago

What error message do you get, please? What version of PyTorch are you using?

csershen commented 6 years ago

Traceback (most recent call last):

File "hello-world.py", line 8, in

import sparseconvnet as scn

File "/brazos/akay/SparseConvNet/SparseConvNet/SparseConvNet/SparseConvNet/sp

arseconvnet/init.py", line 9, in

from .activations import Tanh, Sigmoid, ReLU, LeakyReLU, ELU, SELU,

BatchNo

rmELU

File "/brazos/akay/SparseConvNet/SparseConvNet/SparseConvNet/SparseConvNet/sp

arseconvnet/activations.py", line 11, in

from .utils import *

File "/brazos/akay/SparseConvNet/SparseConvNet/SparseConvNet/SparseConvNet/sp

arseconvnet/utils.py", line 9, in

from .metadata import Metadata

File "/brazos/akay/SparseConvNet/SparseConvNet/SparseConvNet/SparseConvNet/sp

arseconvnet/metadata.py", line 14, in

import sparseconvnet_SCN

ModuleNotFoundError: No module named ‘sparseconvnet_SCN'

using pytorch-0.4.1

Thanks for your reply, Cheryl

On Fri, Aug 17, 2018 at 1:02 AM Ben Graham notifications@github.com wrote:

What error message do you get, please? What version of PyTorch are you using?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/facebookresearch/SparseConvNet/issues/48#issuecomment-413766320, or mute the thread https://github.com/notifications/unsubscribe-auth/AVazLfAquCZ70nQ29NtYexAgolFq-poHks5uRlxagaJpZM4WAoqx .

-- Cheryl Sershen, M.S., M.S., Ph.D.

(530) 902-8251

btgraham commented 6 years ago

I guess the compilation of the sparseconvnet_SCN module must have failed. What happens when you ran bash build.sh?

csershen commented 6 years ago

Hi Ben:

I think I have it running now on a gpu.

I am trying to replicate the Kaggle Diabetic Retinopathy results, do you have a script you could share to help me get started training that model?

Thanks, Cheryl

On Fri, Aug 17, 2018 at 11:27 AM Ben Graham notifications@github.com wrote:

I guess the compilation of the sparseconvnet_SCN module must have failed. What happens when you ran bash build.sh?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/facebookresearch/SparseConvNet/issues/48#issuecomment-413918981, or mute the thread https://github.com/notifications/unsubscribe-auth/AVazLUvMJsg0yAX_SggN38or5MrRo7Guks5uRu7agaJpZM4WAoqx .

-- Cheryl Sershen, M.S., M.S., Ph.D.

(530) 902-8251

csershen commented 6 years ago

PS. I am a researcher at University of Houston and we are doing some research on cancer images and wanted to try your model on those images. Cheryl

On Fri, Aug 17, 2018 at 11:55 AM Cheryl Sershen csershen82@gmail.com wrote:

Hi Ben:

I think I have it running now on a gpu.

I am trying to replicate the Kaggle Diabetic Retinopathy results, do you have a script you could share to help me get started training that model?

Thanks, Cheryl

On Fri, Aug 17, 2018 at 11:27 AM Ben Graham notifications@github.com wrote:

I guess the compilation of the sparseconvnet_SCN module must have failed. What happens when you ran bash build.sh?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/facebookresearch/SparseConvNet/issues/48#issuecomment-413918981, or mute the thread https://github.com/notifications/unsubscribe-auth/AVazLUvMJsg0yAX_SggN38or5MrRo7Guks5uRu7agaJpZM4WAoqx .

-- Cheryl Sershen, M.S., M.S., Ph.D.

(530) 902-8251

-- Cheryl Sershen, M.S., M.S., Ph.D.

(530) 902-8251

btgraham commented 6 years ago

I am afraid that work was done using my previous version https://github.com/btgraham/SparseConvNet, so they cannot be replicated exactly using this version.

What does a typical input look like? Is there anything sparse about your input? If not, I would recommend looking at the models here: https://github.com/pytorch/vision/tree/master/torchvision/models