Helpers utils for manage and track experiments.
Project description
Experiment_utils
Helper utils for track and manage Dl experimets with pytorch.
Very early stage - just draft for my utils.
Install
pip install experiment-utils
Editeble install: git clone https://github.com/ayasyrev/experiment-utils cd experiment-utils pip install -e .
How to use
Import Experiment:
from experiment_utils.experiment import *
After import you has p (stands for Parameters) and e (Experiment) objects.
Name the experiment, later it will be used in logs.
p.exp_name = 'test1'
p.exp_name
'test1'
e.p.exp_name
'test1'
Load learner
e.get_learner()
e.p.model
functools.partial(<function resnet18 at 0x7fc8d4dd08c0>, num_classes=10)
e.learn.model.fc
Linear(in_features=512, out_features=10, bias=True)
Short notation for learn - l
e.l.model.conv1
Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
Now we can easy change some parameter anr start train with pipeline what yuo neg.
from experiment_utils.utils import train_fc, plot
p.pipeline = [train_fc, plot]
p.lr = 0.001
p.epochs = 10
e(repeat_times=2)
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
Hashes for experiment_utils-0.0.3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 4974790c96cd0f052414de1f08ede603126bb813e568e02e4ea433904e3a9f98 |
|
MD5 | c935e56bb0573b21c449783b92f511eb |
|
BLAKE2b-256 | 5ce2d487d82d1b04e360519f5ac58c5305806e968e0636fcd7aee41ab864d744 |