Sign inStart now ›
Training /Track jobs & download results

Jobs & Results

Watch the rounds, read the logs, and download the final model.

Job States

StateWhat is happening
PENDINGThe platform is checking your files and starting the workers.
RUNNINGFL server up, all required workers connected, FL rounds executing.
FINISHEDAll rounds completed — or manually stopped. Output written to files_out/ and model_out/.
FAILEDA worker crashed or timed out. Error message stored on the job. Logs and any files generated by your Python are written to files_out/.
Jobs page — all states visible with worker counts and timestamps

Finding Your Results

When a job reaches FINISHED, output is in two folders in your bucket:

PathContents
jobs/{name}/files_out/Training logs, per-round metrics JSON, any files your train() writes to out_dir. Folders with prefix federated_logs_...date... contain logs from server, worker_logs_...date... - logs from each worker, uuid named folder inside it contains more accurate log.json and log.txt files
jobs/{name}/model_out/Final aggregated model checkpoint after the each FL round (you can download model mid training and check intermediate results)

Option A — from the Job Detail page

  1. 1

    Dashboard → Jobs → click the finished job

  2. 2

    Click "Browse Job Files"

    Opens the Files page pre-navigated to jobs/{name}/.
Job Detail page — Browse Job Files button

The button appears on finished and aborted jobs.

Option B — from the Files page directly

  1. 1

    Dashboard → Files → click your job folder

  2. 2

    Click files_out/ for logs, or model_out/ for the checkpoint

Files page — files_out/ listing with training logs and metrics

Downloading the Final Model

  1. 1

    Navigate to model_out/

  2. 2

    Click the .pt checkpoint file

  3. 3

    Click "Download"

    Download starts directly from S3.
File Viewer — checkpoint.pt with Download button highlighted
note
The checkpoint is a plain PyTorch state_dict. Load it at inference time with model.load_state_dict(torch.load("model_round_N.pt")). No platform dependency required at inference time.
warn
Presigned download URLs expire after 1 hour. If the download times out on a slow connection, click Download again to get a fresh URL.
PreviousDataset format & shardingNext FL examples gallery