nvflare.app_common.tracking.tracker_types module¶
W&B vs MLFLOW vs. Tensorboad DATA –> PARAMETER –> Scalar DATA –> PARAMETERS –> Scalars DATA –> METRIC –> Scalar DATA –> METRICS –> Scalars DATA –> TEXT –> TEXT DATA –> IMAGE –> IMAGE ARTIFACT:type=model –> MODEL
NOT SUPPORTED PLOT –> FIGURE ARTIFACT –> TEXT DATA –> ? –> Histogram ARTIFACT –> DICT ARTIFACT:type=dataset –> ARTIFACT
wandb.log(data) ==> mlflow.log_params(data) ==> writer.add_scalas(data) wandb.log(data) ==> mlflow.log_metrics(data) ==> writer.add_scalas(data) wandb.log({“examples”: images} ==> mlflow.log_image(image, output_path) ==> writer.add_image(‘images’, image, 0)
art = wandb.Artifact(“my-object-detector”, type=”model”) art.add_file(“saved_model_weights.pt”) wandb.log_artifact(art) ==> mlflow.register_model(model_uri, “my-object-detector”) ==> ?
- class TrackConst[source]¶
Bases:
object- DATA_TYPE_KEY = 'analytics_data_type'¶
- EXPERIMENT_NAME = 'experiment_name'¶
- EXPERIMENT_TAG = 'experiment_tag'¶
- EXP_TAGS_KEY = 'tags_key'¶
- GLOBAL_STEP_KEY = 'global_step'¶
- INIT_CONFIG = 'init_config'¶
- JOB_ID_KEY = 'job_id'¶
- KWARGS_KEY = 'analytics_kwargs'¶
- PATH_KEY = 'path'¶
- PROJECT_NAME = 'project_name'¶
- PROJECT_TAGS = 'project_name'¶
- RUN_TAGS = 'run_tags'¶
- SITE_KEY = 'site'¶
- TAGS_KEY = 'tags_key'¶
- TAG_KEY = 'tag_key'¶
- TRACKER_KEY = 'tracker_key'¶
- TRACK_KEY = 'track_key'¶
- TRACK_VALUE = 'track_value'¶