Skip to main content
W&B captures your script’s stdout and stderr and stores it as output.log on the run’s Files tab. By default, that file uploads when the run finishes, so an empty or missing log is often a timing issue, not failed capture. For multipart uploads (console_multipart), chunk rollover, and when to enable them, see Console logs and wandb.Settings. To download logs, see How do I download the console log file from a run?.

​
Console capture is disabled

Console capture can be turned off in settings or with an environment variable: 
wandb.init(settings=wandb.Settings(console="off"))

WANDB_CONSOLE=off python my_script.py
Check whether either is set in your environment or launch configuration. To re-enable, remove the setting or set WANDB_CONSOLE=wrap.

​
Distributed training (DDP / multiprocessing)

The Logs tab records output only from the process that owns the active W&B run. In Lightning/DDP, print() or wandb.termlog() from worker processes that do not own the run appear in the local terminal only. Initialize the run on rank 0 and use console="wrap". 
import wandb
from lightning.pytorch import Trainer
from lightning.pytorch.loggers import WandbLogger

wandb_logger = WandbLogger(
    project="my_project",
    settings=wandb.Settings(console="wrap"),  # or WANDB_CONSOLE=wrap
)
trainer = Trainer(logger=wandb_logger, strategy="ddp", devices=2, accelerator="gpu")
If the Logs tab stays empty, try console="redirect". Output may appear in output.log on the Files tab even when it does not stream live. See Log distributed experiments for rank-0 logging patterns.

​
Run still active but no file on Files tab

While a run is active, the Logs tab streams output, but output.log usually does not appear on Files until the run finishes unless you enabled multipart console logging at wandb.init() time. Upload cadence cannot be changed after the run starts.

​
Run crashed before flush

Without multipart logging, a killed run (OOM, SIGKILL, and similar) may upload no output.log and show no download button. Enable console_multipart before the run starts so chunks uploaded before the crash remain on the server. A local copy is always written to wandb/run-[TIMESTAMP]-[ID]/logs/output.log.

​
Resumed runs lose earlier console output

On older SDKs, wandb.init(resume="allow", id=...) can overwrite a single output.log. With console_multipart=True, each session gets separate chunks under logs/. See Console logs for setup.

​
Logs tab shows fewer lines than expected

The Logs tab caps display for performance: a run stores up to 100,000 lines total, and the App displays up to 10,000 lines at once. Scroll through the logs to view older lines. The full log is in output.log or multipart chunks. Download it from the Files tab of the run details, or through the API.
Logs Runs