I need a PyTorch Programmer
Budget: €200 – €300 EUR
I need a capable pytorch programmer who is able to develop a pytorch based RNN, more precisely a GRU which is able to be used in the framework "Pytorch-Forecasting". The GRU must be able to receive the dataloaders of the framwork's own TimeSeriesDataset. The framework is based on PyTorch and already has a GRU integrated. The problem is that the GRU cannot process covariates. However, I have several covariates and therefore need a GRU that can handle the dataloader.
The problem is described here for example: https://github.com/jdb78/pytorch-forecasting/issues/553
The cause is described by someone like this:
https://github.com/jdb78/pytorch-forecasting/issues/590
According to the docs it is possible to implement your own models: https://pytorch-forecasting.readthedocs.io/en/stable/tutorials/building.html
At the place is also described how the content of the Dataloader looks like.
I need a GRU which gets the dataloader from the TimeSeriesDataSet and can be trained on it and make predictions and plot them.
The data is already in a dataframe with the correct structure for the TimeSeriesDataSet, namely with a "Series" and a "time_idx" column.
A series consists of 36 timesteps where the first 30 time steps are to be used as input while the remaining 6 time steps are to be predicted.
It is very important that the GRU is able to train with the data from the framework own dataloader and use the covariates.
Please have a look at the things in the links and see if you are able to solve this.
Contact me so i can show you the data.
The problem is described here for example: https://github.com/jdb78/pytorch-forecasting/issues/553
The cause is described by someone like this:
https://github.com/jdb78/pytorch-forecasting/issues/590
According to the docs it is possible to implement your own models: https://pytorch-forecasting.readthedocs.io/en/stable/tutorials/building.html
At the place is also described how the content of the Dataloader looks like.
I need a GRU which gets the dataloader from the TimeSeriesDataSet and can be trained on it and make predictions and plot them.
The data is already in a dataframe with the correct structure for the TimeSeriesDataSet, namely with a "Series" and a "time_idx" column.
A series consists of 36 timesteps where the first 30 time steps are to be used as input while the remaining 6 time steps are to be predicted.
It is very important that the GRU is able to train with the data from the framework own dataloader and use the covariates.
Please have a look at the things in the links and see if you are able to solve this.
Contact me so i can show you the data.