Saving Files

Overview

W&B saves configuration parameters, custom & system metrics, along with standard out / error by default. W&B can also save arbitrary files associated with a run. This is especially useful for saving the literal weights and biases.

Example

import wandb
wandb.init()

model.fit(X_train, y_train,  validation_data=(X_test, y_test),
    callbacks=[wandb.keras.WandbCallback()])
model.save(os.path.join(wandb.run.dir, "model.h5")) #

W&B will save to the cloud any files put in wandb's run directory. If you are already saving a file in a different directory and want it synced to wandb, you can run the following to create a symlink and ensure syncing.

wandb.save('path/to/existing/file.pb', overwrite=True)

W&B's local run directories are by default inside the ./wandb directory relative to your script, and the path looks like run-20171023_105053-3o4933r0 where 20171023_105053 is the timestamp and 3o4933r0 is the ID of the run. You can set the WANDB_DIR environment variable, or the dir keyword argument of wandb.init to an absolute path and files will be written within that directory instead.

Ignoring certain files

You can edit the wandb/settings file and set ignore_globs equal to a comma seperated list of globs. You can also set the WANDB_IGNORE_GLOBS environment variable. A common usecase is to prevent the git patch that we automatically create from being uploaded i.e. WANDB_IGNORE_GLOBS=diff.patch