Open BaoxiJia opened 2 years ago
@BaoxiJia We fixed this in the latest release. Did you use 0.38.1 installer?
@penorouzi I used 0.38.0 for testing.
@BaoxiJia Can you use the latest version and see if you still have the issue?
I tried the latest 0.38.1, the updated model couldn't take effect until I rebooted PredictModule manually on Edge side. Here is what I found -
What I found:
["skill-13580f31-719e-45e5-90ec-2347c6c54aef as skill-d92d"]
parameters={'configure_data': '{"cam-cpu": ["car-skill-cpu"]}', 'skill-d92d.device_displayname': 'cam-cpu', 'skill-d92d.device_id': 'device-96913022-58a8-40b0-872a-121cc5f78ffc', 'skill-d92d.fps': '15', 'skill-d92d.instance_displayname': 'deploy-car-cpu', 'skill-d92d.rtsp': 'rtsp://:@10.172.88.62/media/camera-300s.mkv', 'skill-d92d.skill_displayname': 'car-skill-cpu'} solution='solution-01cc9761-2fc6-4947-a3fe-72c4058724e4' target=InstanceTarget(name='target-cb6f2bb4-882f-475d-aee9-94e5083e5f5b')
ok
skill_name: skill-13580f31-719e-45e5-90ec-2347c6c54aef, skill_alias: skill-d92d
{'apiVersion': 'ai.symphony/v1', 'kind': 'Skill', 'metadata': {'creationTimestamp': '2022-09-02T04:56:10Z', 'generation': 9, 'name': 'skill-13580f31-719e-45e5-90ec-2347c6c54aef', 'namespace': 'default', 'resourceVersion': '6810096', 'uid': 'dcc1893c-5f60-4232-b476-6e62cb9e9ef2'}, 'spec': {'displayName': 'car-skill-cpu', 'edges': [{'source': {'node': '0', 'route': 'f'}, 'target': {'node': '1', 'route': 'f'}}, {'source': {'node': '1', 'route': 'f'}, 'target': {'node': '2', 'route': 'f'}}, {'source': {'node': '1', 'route': 'f'}, 'target': {'node': '3', 'route': 'f'}}], 'nodes': [{'configurations': {'device_name': 'device-96913022-58a8-40b0-872a-121cc5f78ffc', 'fps': '15', 'ip': 'rtsp://:@10.172.88.62/media/camera-300s.mkv'}, 'id': '0', 'name': 'rtsp', 'type': 'source'}, {'configurations': {'confidence_lower': '0', 'confidence_upper': '0', 'max_images': '0'}, 'id': '1', 'name': 'model-c1ff2999-2c53-48b8-9db6-b9a9dfe0bc0e', 'type': 'model'}, {'configurations': {'delay_buffer': '2', 'device_displayname': 'cam-cpu', 'filename_prefix': 'hcicpu', 'insights_overlay': 'true', 'instance_displayname': 'deploy-car-cpu', 'recording_duration': '10', 'skill_displayname': 'car-skill-cpu'}, 'id': '2', 'name': 'video_snippet_export', 'type': 'export'}, {'configurations': {'delay_buffer': '30'}, 'id': '3', 'name': 'iothub_export', 'type': 'export'}], 'parameters': {'accelerationRetrieve': 'CPU', 'device_displayname': 'invalid', 'device_id': 'invalid', 'fps': 'invalid', 'fpsRetrieve': '15', 'instance_displayname': 'invalid', 'rtsp': 'invalid', 'skill_displayname': 'invalid'}}}
s --> properties=ModelProperties(model_type='customvision', model_subtype='customvision.ObjectDetection', model_project='392db6ef-966c-46a5-964f-f0845fe72e3f', tags='["person", "car"]', state='trained')
Got Download URI https://irisprodwu2training.blob.core.windows.net:443/m-392db6ef966c46a5964ff0845fe72e3f/cef2a7edbe004e1e9e1230794d1153d0.ONNX.zip?sv=2020-04-08&se=2022-09-03T06%3A04%3A32Z&sr=b&sp=r&sig=z5RhScLjPw86aKyKSUY34CHreSIJ7%2FqvsjP78Vey8nI%3D
------------------------------------------------------------
Getting status from StreamingModule and PredictModule ...
------------------------------------------------------------
Predict Module: StatusEnum.RUNNING
Streaming Module: StatusEnum.WAITING
------------------------------------------------------------
sending cascade_configs to predictmodule ...
------------------------------------------------------------
[CascadeConfig(edges=[Edge(source='0', target='1'), Edge(source='1', target='2'), Edge(source='1', target='3')], nodes=[Node(id='0', type='source', name='rtsp', configurations={'device_name': 'device-96913022-58a8-40b0-872a-121cc5f78ffc', 'fps': '15', 'ip': 'rtsp://:@10.172.88.62/media/camera-300s.mkv', 'skill_name': 'skill-13580f31-719e-45e5-90ec-2347c6c54aef'}), Node(id='1', type='model', name='object_detection_model', configurations={'confidence_lower': '0', 'confidence_upper': '0', 'max_images': '0', 'model': 'model-c1ff2999-2c53-48b8-9db6-b9a9dfe0bc0e', 'symphony_name': 'model-c1ff2999-2c53-48b8-9db6-b9a9dfe0bc0e', 'provider': 'customvision'}), Node(id='2', type='export', name='video_snippet_export', configurations={'delay_buffer': '2', 'device_displayname': 'cam-cpu', 'filename_prefix': 'hcicpu', 'insights_overlay': 'true', 'instance_displayname': 'deploy-car-cpu', 'recording_duration': '10', 'skill_displayname': 'car-skill-cpu'}), Node(id='3', type='export', name='iothub_export', configurations={'delay_buffer': '30'})])]
INFO: 172.18.0.7:48194 - "POST /predict/model-c1ff2999-2c53-48b8-9db6-b9a9dfe0bc0e?width=300&height=300 HTTP/1.1" 200 OK
INFO: 172.18.0.7:51008 - "POST /predict/model-c1ff2999-2c53-48b8-9db6-b9a9dfe0bc0e?width=300&height=300 HTTP/1.1" 200 OK
2022-09-02 06:28:04.788613241 [W:onnxruntime:, execution_frame.cc:811 VerifyOutputSizes] Expected shape from model of {None,35,13,13} does not match actual shape of {1,35,16,16} for output model_outputs0
--> modes {'model-c1ff2999-2c53-48b8-9db6-b9a9dfe0bc0e': <customvision_object_detection.CustomVisionObjectDetectionModel object at 0x78fec76d7bb0>}
--> model-c1ff2999-2c53-48b8-9db6-b9a9dfe0bc0e
Thank you @BaoxiJia for the detailed info.
@ronpai @waitingkuo Please look at the above issue.
Check the deployment running on the Edge device, I can see it still inferences using Model A, while Model B is expected. To make the updated AI skill work, I have to delete the current deployment and recreate it with the updated AI skill.