Reduce the logging/verbosity of orbax async checkpointing

[zaccharieramzi]

I am currently using orbax to save/load my model checkpoints as per the official docs.
However, I am faced with a lot of logs since I am using the async checkpointer:

try:
    jax.distributed.initialize("localhost:8889", num_processes=1, process_id=0)
except RuntimeError:
    pass
orbax_checkpointer = orbax.checkpoint.AsyncCheckpointer(
    orbax.checkpoint.PyTreeCheckpointHandler(),
    timeout_secs=50,
)
options = orbax.checkpoint.CheckpointManagerOptions(
    max_to_keep=2,
    create=True,
)
checkpoint_manager = orbax.checkpoint.CheckpointManager(
    'checkpoints',
    orbax_checkpointer,
    options,
)

...
# training loop

ckpt = {
    "params": params,
    "state": state,
    "batch_stats": batch_stats,
    "cfg": cfg,
}
save_args = orbax_utils.save_args_from_target(ckpt)
checkpoint_manager.save(i, ckpt, save_kwargs={'save_args': save_args})

I would like to know how to reduce the amount of logs/the verbosity level of the saving.
I haven't seen how in the docs.
I am posting here, since I did not see a discussion tab in the orbax repo.

[spapinissi]

I solved my issue by reducing log leve to info just before loading orbax checkpoint:

import absl.logging
absl.logging.set_verbosity(absl.logging.INFO)

[evgenytsydenov]

I'm afraid this will not help since orbax creates all logs as INFO messages:

I1108 19:20:54.522927 140266138764992 async_checkpointer.py:168] [process=0][thread=async_save_1] Background save thread done.
I1108 19:20:54.523293 140266004547264 async_checkpointer.py:226] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save_1.
I1108 19:20:54.523516 140266004547264 async_checkpointer.py:236] [process=0][thread=save_finalize] No errors found in background save thread=async_save_1.
...