Documentation
Search…
MMF
The WandbLogger class in Meta AI's MMF library will enable Weights & Biases to log the training/validation metrics, system (GPU and CPU) metrics, model checkpoints and configuration parameters.

Current features

The following features are currently supported by the WandbLogger in MMF:
  • Training & Validation metrics
  • Learning Rate over time
  • Model Checkpoint saving to W&B Artifacts
  • GPU and CPU system metrics
  • Training configuration parameters

Config parameters

The following options are available in MMF config to enable and customize the wandb logging:
1
training:
2
wandb:
3
enabled: true
4
5
# An entity is a username or team name where you're sending runs.
6
# By default it will log the run to your user account.
7
entity: null
8
9
# Project name to be used while logging the experiment with wandb
10
project: mmf
11
12
# Experiment/ run name to be used while logging the experiment
13
# under the project with wandb. The default experiment name
14
# is: ${training.experiment_name}
15
name: ${training.experiment_name}
16
17
# Turn on model checkpointing, saving checkpoints to W&B Artifacts
18
log_model_checkpoint: true
19
20
# Additional argument values that you want to pass to wandb.init().
21
# Check out the documentation at https://docs.wandb.ai/ref/python/init
22
# to see what arguments are available, such as:
23
# job_type: 'train'
24
# tags: ['tag1', 'tag2']
25
26
env:
27
# To change the path to the directory where wandb metadata would be
28
# stored (Default: env.log_dir):
29
wandb_logdir: ${env:MMF_WANDB_LOGDIR,}
Copied!