Erick Fonseca
Erick Fonseca
If your GPU memory is small, the only solution I can think of is running a smaller batch size.
That sounds very weird. I don't have any other suggestions, as I haven't touched this code for over a year now. Which version of Tensorflow are you using? This code...
This is likely a problem with your hyperparameters, since even the training loss isn't improving. I would guess you are using a very high learning rate and/or maybe a high...
Ah, see that the `-d` parameter actually means the dropout *keep* probability, so 0.2 means you are zeroing 80% of the inputs. You probably meant to use `-d 0.8`. It...
This is strange. I consistently get good results with the configuration mentioned in the readme.
Sorry for the delay, I haven't had much time to look at nlpnet, but I'll try to check it this week.
Thanks for the reply. I thought about converting the batch to half precision as well, the problem is that the batch contents are a list of PIL images, and the...
Ok, but now I realize that this doesn't really test if Yolo correctly interpreted the fp16 weights as such. If I change the inputs to half precision, the values will...
I loaded them with the `transformers` library. For Czech it was `Helsinki-NLP/opus-mt-de-cs` and for English `Helsinki-NLP/opus-mt-de-en`.
Just install the overrides module with `pip install overrides`. It's a very simple module that provides a decorator for basic checking overriding functions.