CUDA Driver version problem.

[androidgeek077]

When i try to train the the birds model, I encountered this problem.

CUDA driver version is insufficient for CUDA runtime version at torch/csrc/cuda/Module.cpp:32

[MinfengZhu]

Have you installed the cudatoolkit=9.0?
cudatoolkit=9.0 may be incompatible in your environment.

conda install requests nltk pandas scikit-image pyyaml cudatoolkit=9.0

[androidgeek077]

Hello MinfengZhu ! Thanks for replying sir. I figured out that i didn;t have CUDA GPUs. Can you please tell me that I can run this code without CUDA GPU? I'm a student of masters working on your bases as my base paper. I shall be very thankful to you. Regards

…

On Fri, 17 Jul 2020 at 23:54, TwilightSnow ***@***.***> wrote: Have you installed the cudatoolkit=9.0? cudatoolkit=9.0 may be incompatible in your environment. conda install requests nltk pandas scikit-image pyyaml cudatoolkit=9.0 — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AL4PEROLKCPE6GTUBFY5AQTR4FBJFANCNFSM4OKOLWSQ> .

[MinfengZhu]

This code should work on CPU when you set --gpu=-1. Besides, please carefully check that all variables are not moved to GPUs. You can remove all .cuda() commands.

[hs457681503]

When i try to train the the birds model, I encountered this problem.

Coulde you help me solve this problem?

RuntimeError: Error(s) in loading state_dict for RNN_ENCODER:
size mismatch for encoder.weight: copying a param of torch.Size([1, 300]) from checkpoint, where the shape is torch.Size([5450, 300]) in current model.

[MinfengZhu]

When i try to train the the birds model, I encountered this problem.

Coulde you help me solve this problem?

RuntimeError: Error(s) in loading state_dict for RNN_ENCODER:
size mismatch for encoder.weight: copying a param of torch.Size([1, 300]) from checkpoint, where the shape is torch.Size([5450, 300]) in current model.

Could you please provide more information (e.g., config file or command)? If you train the model from scratch, there is no checkpoint to load.

[savitha91]

Hi I am using Colab to execute the code. I am unable to install cudatoolkit=9.0 and the pytorch installation given. I am using pre-trained model, so when i run the command ! python main.py --cfg eval_coco.yml --gpu 0 . i get the error , No such file or directory: u'../data/coco/captions.pickle' . I have just places examples_filenames.txt into the data/coco folder. Nothing is mentioned about the captions.pickle

[androidgeek077]

I am using example_filename.txt and generating images using main.py. But I am achieving IS 2.92 only. Can you please guide me why I am not getting the same mentioned IS score as you have mentioned in your paper. For FID, I am getting 65.48. Please guide me. I have also mentioned it in issues Thanks and Regards

…

On Mon, 3 Aug 2020 at 09:06, savitha91 ***@***.***> wrote: Hi I am using Colab to execute the code. I am unable to install cudatoolkit=9.0 and the pytorch installation given. I am using pre-trained model, so when i run the command ! python main.py --cfg eval_coco.yml --gpu 0 . i get the error , No such file or directory: u'../data/coco/captions.pickle' . I have just places examples_filenames.txt into the data/coco folder. Nothing is mentioned about the captions.pickle — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AL4PERMEFTBA7PK3W5Z242TR6YZUXANCNFSM4OKOLWSQ> .

[MinfengZhu]

Hi I am using Colab to execute the code. I am unable to install cudatoolkit=9.0 and the pytorch installation given. I am using pre-trained model, so when i run the command ! python main.py --cfg eval_coco.yml --gpu 0 . i get the error , No such file or directory: u'../data/coco/captions.pickle' . I have just places examples_filenames.txt into the data/coco folder. Nothing is mentioned about the captions.pickle

I hope DM-GAN can generate images with consistent quality using different pytorch versions.

Have you downloaded the metadata for COCO dataset?
python google_drive.py 1rSnbIGNDGZeHlsUlLdahj0RJ9oo6lgH9 ./data/coco.zip

[MinfengZhu]

I am using example_filename.txt and generating images using main.py. But I am achieving IS 2.92 only. Can you please guide me why I am not getting the same mentioned IS score as you have mentioned in your paper. For FID, I am getting 65.48. Please guide me. I have also mentioned it in issues Thanks and Regards
…

Please generate images using the following command. You should generate images using text descriptions from the whole valid dataset.
python main.py --cfg cfg/eval_bird.yml --gpu 0

[androidgeek077]

Thanks Sir ! Let me try this one

…

On Wed, 5 Aug 2020 at 17:45, TwilightSnow ***@***.***> wrote: I am using example_filename.txt and generating images using main.py. But I am achieving IS 2.92 only. Can you please guide me why I am not getting the same mentioned IS score as you have mentioned in your paper. For FID, I am getting 65.48. Please guide me. I have also mentioned it in issues Thanks and Regards … <#m_2445682859946621279_> Please generate images using the following command. You should generate images using text descriptions from the whole valid dataset. python main.py --cfg cfg/eval_bird.yml --gpu 0 — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AL4PERLZFFKTIEZTYCLGLD3R7FH7HANCNFSM4OKOLWSQ> .

[androidgeek077]

Hello Sir ! its me again looking for a little favor. in the performance section of your repository's readme file you have mentioned that " *after cleaning and refactoring the code of the paper, the results are slightly different* " I just want to know what type of refactoring or cleaning you have done to get even better results then the mentioned one in paper. If you have any document related to this then it will be great otherwise a brief explanation will be enough. I shall be very thankful to you if you do me this favor. This is my masters thesis. DM-GAN is my base paper. I'm heavily replying on this model. Kind regards Waqas Ahmad

…

On Thu, 6 Aug 2020 at 00:28, android geek ***@***.***> wrote: Thanks Sir ! Let me try this one On Wed, 5 Aug 2020 at 17:45, TwilightSnow ***@***.***> wrote: > I am using example_filename.txt and generating images using main.py. But > I am achieving IS 2.92 only. Can you please guide me why I am not getting > the same mentioned IS score as you have mentioned in your paper. For FID, I > am getting 65.48. Please guide me. I have also mentioned it in issues > Thanks and Regards > … <#m_-2588125760430073018_m_2445682859946621279_> > > Please generate images using the following command. You should generate > images using text descriptions from the whole valid dataset. > python main.py --cfg cfg/eval_bird.yml --gpu 0 > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#14 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AL4PERLZFFKTIEZTYCLGLD3R7FH7HANCNFSM4OKOLWSQ> > . >

[MinfengZhu]

We train DM-GAN with much more epochs (please see yml file in cfg folder) and choose the epoch which has the best performance.

We report the performance on CUB dataset using 600 epochs in our paper. The released model on CUB is trained using 800 epochs (bird_DMGAN.yml).

[cloverzyy]

When i try to train the the birds model, I encountered this problem.

Coulde you help me solve this problem?

RuntimeError: Error(s) in loading state_dict for RNN_ENCODER:
size mismatch for encoder.weight: copying a param of torch.Size([1, 300]) from checkpoint, where the shape is torch.Size([5450, 300]) in current model.

Did you solve the problem? I have the same problem. Could you tell me how to modify the code or add something? Thank you!

[MinfengZhu]

When i try to train the the birds model, I encountered this problem.
Coulde you help me solve this problem?
RuntimeError: Error(s) in loading state_dict for RNN_ENCODER:
size mismatch for encoder.weight: copying a param of torch.Size([1, 300]) from checkpoint, where the shape is torch.Size([5450, 300]) in current model.

Did you solve the problem? I have the same problem. Could you tell me how to modify the code or add something? Thank you!

Have you correctly load pertrained models? Text encoder and image encoder are already provided in DAMSM.

[srinivaspavan9]

This code should work on CPU when you set --gpu=-1. Besides, please carefully check that all variables are not moved to GPUs. You can remove all .cuda() commands.

Can you please elaborate this on how to make it work for cpus with out CUDA. I am still getting the same error.

[srinivaspavan9]

Hello MinfengZhu ! Thanks for replying sir. I figured out that i didn;t have CUDA GPUs. Can you please tell me that I can run this code without CUDA GPU? I'm a student of masters working on your bases as my base paper. I shall be very thankful to you. Regards
…
On Fri, 17 Jul 2020 at 23:54, TwilightSnow @.***> wrote: Have you installed the cudatoolkit=9.0? cudatoolkit=9.0 may be incompatible in your environment. conda install requests nltk pandas scikit-image pyyaml cudatoolkit=9.0 — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AL4PEROLKCPE6GTUBFY5AQTR4FBJFANCNFSM4OKOLWSQ .

I am facing the same issue, can you help me on how to execute on cpu with out cuda.