run

Start the training through API.

post
Start Training

http://localhost:1004/api/v1/run
This endpoint allows you to start the training.
Request
Response
Request
Body Parameters
TRAINER_CONFIG
required
object
JSON containing configuration required to start the training.
Response
200: OK
Training has started successfully.
{"success": "flow started"}
404: Not Found
Could not find the flow file.
{"error":"/app/flows/flow_SSd.py not found"}

Example

{ "api_key":"<API-KEY>",
"images": "/dataset/images",
"annotations": "/dataset/annotations",
"resize":{"width": 300, "height": 300, "padding":true},
"classes": ["face", "person"],
"pretrained_model":"/pretrained/ssd_resnet10_epoch_5000.tlt",
"architecture":{
"name": "DetectNet_v2",
"backbone":"ResNet10",
"batch_size":32,
"epochs":2
}
}

Note:-

  • {"images": "/dataset/images"} Here the path is indicating mounted images path inside the container and not on the host machine.

  • {"pretrained_model":"/pretrained/ssd_resnet10_epoch_5000.tlt"} This parameter is optional, and if not included in the JSON body then the trainer will download base weights from NGC according to the backbone.