Parameters (**kargs)
A splecial class whos atributes can be referenced as attributs or as dictionaty keys
device_by_name (name:str)
Return reference to cuda device by using Part of it’s name
Args: name: part of the cuda device name (shuuld be distinct)
Return: Reference to cuda device
Updated: Yuval 12/10/19
DatasetCat (*datasets)
Concatenate datasets for Pytorch dataloader
The normal pytorch implementation does it only for raws. this is a “column” implementation
Arges: datasets: list of datasets, of the same length
Updated: Yuval 12/10/2019
pip install reinautilsYou can create a Parameters class from dict
params=Parameters(first=1,second='A')
print(params.first)1
You can also creat a Parameters class and populate it from a json file
params2=Parameters().from_json('config_demo.json')
print(params2)Parameters:
path : Parameters:
data : /workspace/hd/
tmp : /workspace/hd/tmp/
features : /workspace/nvme/features/
train : /workspace/nvme/train/
models : /workspace/hd/models/
output : /workspace/hd/outputs/
test : /workspace/nvme/test/
platform : myserver