update train

supervisely/train/src/sly_train_globals.py

@@ -3,6 +3,7 @@
 import sys
 import yaml
 import supervisely as sly
+from supervisely.nn.checkpoints.yolov5 import YOLOv5Checkpoint
 from supervisely.app.v1.app_service import AppService
 from dotenv import load_dotenv
 
@@ -54,6 +55,14 @@
 experiment_name = str(task_id)
 local_artifacts_dir = os.path.join(runs_dir, experiment_name)
 sly.logger.info(f"All training artifacts will be saved to local directory {local_artifacts_dir}")
-remote_artifacts_dir = os.path.join("/yolov5_train", project_info.name, experiment_name)
+
+checkpoint = YOLOv5Checkpoint(team_id)
+model_dir = checkpoint.get_model_dir()
+
+remote_artifacts_dir = os.path.join(model_dir, project_info.name, experiment_name)
 remote_artifacts_dir = api.file.get_free_dir_name(team_id, remote_artifacts_dir)
+
+remote_weights_dir = os.path.join(remote_artifacts_dir, checkpoint.weights_dir)
+remote_weights_dir = api.file.get_free_dir_name(team_id, remote_artifacts_dir)
+
 sly.logger.info(f"After training artifacts will be uploaded to Team Files: {remote_artifacts_dir}")

supervisely/train/src/sly_utils.py

@@ -66,4 +66,15 @@ def _gen_message(current, total):
             globals.api.file.upload(globals.team_id, local_path, remote_path,
                                     lambda monitor: progress_cb(progress_last + monitor.bytes_read))
         progress.message = _gen_message(idx + 1, len(local_files))
-        time.sleep(0.5)
+        time.sleep(0.5)
+
+    # generate metadata
+    globals.checkpoint.generate_sly_metadata(
+        app_name=globals.checkpoint.app_name,
+        session_id=globals.experiment_name,
+        session_path=globals.remote_artifacts_dir,
+        weights_dir=globals.remote_weights_dir,
+        training_project_name=globals.project_info.name,
+        task_type=globals.checkpoint.task_type,
+        config_path=None,
+    )