Select module classes and functions using yaml, without any if-statements.
Project description
easy_module_attribute_getter
Installation:
pip install easy_module_attribute_getter
Simple example, using PytorchGetter which extends EasyModuleAttributeGetter:
1. Specify class names and arguments in your yaml config file:
models:
modelA:
densenet121:
pretrained: True
memory_efficient: True
modelB:
resnext50_32x4d:
pretrained: True
losses:
lossA:
CrossEntropyLoss:
lossB:
L1Loss:
2. Read yaml file, and get objects from modules:
from easy_module_attribute_getter import YamlReader, PytorchGetter
yaml_reader = YamlReader()
args, _, _ = yaml_reader.load_yamls(['example.yaml'])
pytorch_getter = PytorchGetter()
models = pytorch_getter.get_multiple("model", args.models)
losses = pytorch_getter.get_multiple("loss", args.losses)
"models" is a dictionary with keys "modelA" and "modelB" as specified in the yaml file. The values are the loaded objects, which in this case are PyTorch models. The same is true for "losses".
Easily register your own modules into an existing getter.
from pytorch_metric_learning import losses, miners, samplers
pytorch_getter = PytorchGetter()
pytorch_getter.register('loss', losses)
pytorch_getter.register('miner', miners)
pytorch_getter.register('sampler', samplers)
metric_loss = pytorch_getter.get('loss', class_name='TripletMarginLoss', return_uninitialized=True)
kl_div_loss = pytorch_getter.get('loss', class_name='KLDivLoss', return_uninitialized=True)
In the above example, the 'loss' key already exists, so the 'losses' module will be appended to the existing module. Both modules are searched when get() or get_multiple() is used. For example, metric_loss contains an object from the pytorch_metric_learning library, whil kl_div_loss contains an object from the PyTorch library, which was registered by default.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Close
Hashes for easy_module_attribute_getter-0.9.13.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | c61b7055b17b24021c13717b21883217f0c74c60f72bfdc9026f3aa15df4ca43 |
|
MD5 | b704450890ee44dea71e4dfafee2dba9 |
|
BLAKE2b-256 | f77fe536e3e4412c3186da07009d8566fb15fd526b8a9c6f1495cc3d6f501b2e |
Close
Hashes for easy_module_attribute_getter-0.9.13-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | fdd3e9cf7ec39f19565245c72e364f6afb305deee4ad72eeaeee3c8d7a90b6a8 |
|
MD5 | 268a6977d10ce8eba1d00a30d5bd210b |
|
BLAKE2b-256 | 059f4000990b5cc62b4a31189b813952bcce770aa5e7f0096358d79cd436ee44 |