Mariam

Okay, thanks a lot. I did the text-image generation (synthetic) used in this project (with about 5 different fonts and 5 different text polices but with a very big japanese...

Yes but the composition isn't very varient so I'm worried probably it won't recognize anything I give it to it later (disregarding the accuracy of the model I mean).

Yes, by composition I mean for example I create images with 10 different backgrounds instead of 5 different bakgrounds which mean the model will learn different background features. That's the...

I see, for me I used a text file that has a huge Japanese and English text corpus (adresses, cities, names, dates, ...etc): its size is 60Mo (the file) and...

In the code there's already: gpu number (params.ngpu): crnn = torch.nn.DataParallel(crnn, device_ids=range(params.ngpu)) Do I change it to something like cuda:0, cuda:1,...? like that?

Alright, I will check it, but can you check out this link about gpus parallalesim? https://discuss.pytorch.org/t/multi-gpu-training-pipeline-in-0-4-1/32199 To recap what's written in this link: when we have an optimizer, the gpus...

Alright, I got the following error: AttributeError: 'RMSprop' object has no attribute 'cuda'

I haven't tested it yet too, but the error appears before the execution: the optimizer doesn't have cuda in it. I thought about deleting the optimizers (lol).

Now I erased the optimizer (just to see the result) and the error is about: RNN cannot be divided and that I should add the flatten_parameters function to it.

Thanks for the answer, I will try it. So the problem is about the dimensions and not about the optimizer? In models/crnn.py file, do we change: self.rnn = nn.LSTM(nIn, nHidden,...