Skip to content

Latest commit

 

History

History
118 lines (83 loc) · 3.83 KB

config.md

File metadata and controls

118 lines (83 loc) · 3.83 KB
description
Dictionary-like object to save your experiment configuration

wandb.config

Overview

Use wandb.config to save your training config: hyperparameters, input settings like dataset name or model type, and any other independent variables for your experiments. This is useful for analyzing your experiments and reproducing your work in the future. You'll be able to group by config values in the web interface, comparing the settings of different runs and seeing how these affect the output. Note that output metrics or dependent variables (like loss and accuracy) should be saved with wandb.loginstead. Config should be set just once at the beginning of your training experiment.

You can send us a nested dictionary in config, and we'll flatten the names using dots in our backend. We recommend that you avoid using dots in your config variable names, and use a dash or underscore instead. Once you've created your wandb config dictionary, if your script accesses wandb.config keys below the root, use [ ] syntax instead of . syntax.

Simple Example

wandb.config.epochs = 4
wandb.config.batch_size = 32
# you can also initialize your run with a config
wandb.init(config={"epochs": 4})

Efficient Initialization

You can treat wandb.config as a dictionary, updating multiple values at a time.

wandb.init(config={"epochs": 4, "batch_size": 32})
# or
wandb.config.update({"epochs": 4, "batch_size": 32})

TensorFlow Flags

You can pass TensorFlow flags into the config object.

wandb.init()
wandb.config.epochs = 4

flags = tf.app.flags
flags.DEFINE_string('data_dir', '/tmp/data')
flags.DEFINE_integer('batch_size', 128, 'Batch size.')
wandb.config.update(flags.FLAGS)  # adds all of the tensorflow flags as config

Argparse Flags

You can pass in the arguments dictionary from argparse. This is convenient for quickly testing different hyperparameter values from the command line.

wandb.init()
wandb.config.epochs = 4

parser = argparse.ArgumentParser()
parser.add_argument('-b', '--batch-size', type=int, default=8, metavar='N',
                     help='input batch size for training (default: 8)')
args = parser.parse_args()
wandb.config.update(args) # adds all of the arguments as config variables

File-Based Configs

You can create a file called config-defaults.yaml, __and it will automatically be loaded into wandb.config

# sample config defaults file
epochs:
  desc: Number of epochs to train over
  value: 100
batch_size:
  desc: Size of each mini-batch
  value: 32

You can tell wandb to load different config files with the command line argument --configs special-configs.yaml which will load parameters from the file special-configs.yaml.

One example use case: you have a YAML file with some metadata for the run, and then a dictionary of hyperparameters in your Python script. You can save both in the nested config object:

hyperparameter_defaults = dict(
    dropout = 0.5,
    batch_size = 100,
    learning_rate = 0.001,
    )

config_dictionary = dict(
    yaml=my_yaml_file,
    params=hyperparameter_defaults,
    )
    
wandb.init(config=config_dictionary)

Dataset Identifier

You can add a unique identifier (like a hash or other identifier) in your run's configuration for your dataset by tracking it as input to your experiment using wandb.config

wandb.config.update({'dataset':'ab131'})

Update Config Files

You can use the public API to update your config file

import wandb
api = wandb.Api()
run = api.run("username/project/run_id")
run.config["foo"] = 32
run.update()

Key Val Pairs

ou can log any key val pairs into wandb.config. They will be different for every type of model you are training. i.e. wandb.config.update({"my_param": 10, "learning_rate": 0.3, "model_architecture": "B"})