2 years ago
I'm developing an API in Django Rest Framework, and in one of my applications called 'Predict,' I've created a directory named 'data' where I have a file handling the logic to obtain data and pass it to a pre-trained model 'Model.joblib,' which is in the same directory as the file. Now, on my local machine, it works by passing the absolute path, but in production on 'Railway,' it doesn't work with either absolute or relative paths.
I have no idea what it could be, I haven't investigated thoroughly but I really need to fix this immediately, I appreciate your help in advance.
25 Replies
2 years ago
Can you share some of the paths you've tried, what paths work locally and what paths don't work on Railway, and the errors you are getting.
2 years ago
There is the implementation locally and the initial error.
Attachments
2 years ago
Well, as I mentioned earlier, I superficially investigated, and it was suggested to me to check the issue with paths. Still, I continue testing, and if I find something, I will document it
2 years ago
This happens when I use an absolute path.
"Error al sincronizar el dataset:Layer 'simplernncell' expected 3 variables, but received 0 variables during loading. Expected: ['simplernn8/simplernncell/kernel:0', 'simplernn8/simplernncell/recurrentkernel:0', 'simplernn8/simplernn_cell/bias:0']"
Attachments
2 years ago
English only, please.
Without knowing your project structure or your deployment strategy I would just be guessing, So I would recommend just keep debugging this, different paths, etc.
2 years ago
It's important to emphasize that this exception doesn't occur locally.
2 years ago
If you are using a production environment, make sure that access permissions to the model and storage directories are correct and properly configured.
Could this be a reason? Since I consider Railway to be quite pragmatic with file exposure, in my opinion.
2 years ago
I understand, so I would have to validate the model construction itself to rule that out as well. Thank you very much for the clarifications and help.
2 years ago
Just had a thought, is the model in Git LFS? If so, Railway does not pull files from Git LFS.
2 years ago
The model is from TensorFlow, it's in '.joblib' format, and this is where I compiled it, so to speak.
from joblib import dump
dump(bestrfmodel, 'model.joblib')
2 years ago
No, the model is in its original extension, I just verified it, or at least I think so, but yes, I uploaded the binary to the repository
2 years ago
Well, I wouldn't know exactly when GitHub or Git does it through Git LFS, so I would believe it's not that type of file. How could I check it? I think I didn't attach a screenshot in my previous response.
I think the attachment limit has been exceeded.
2 years ago
No worries if you don't know what Git LFS is or what it entails, that's why GitHub writes documentation!
2 years ago
PS C:\ProyectosSChang\VictoriaAPI> git lfs ls-files PS C:\ProyectosSChang\VictoriaAPI>
PS C:\ProyectosSChang\VictoriaAPI> git lfs ls-files PS C:\ProyectosSChang\VictoriaAPI>
I think I have ruled out that possibility
2 years ago
144 kb
2 years ago
That's right, I will validate that tomorrow, I will keep you informed