Tags

Version generated on 23/06/29

Fixed issues with the maximum workload API; removed temporary fix
used to deal with wrong unit of measures for the throughput.
Tested on:
- blurry_faces_single_component_local_constraint
from https://gitlab.polimi.it/ai-sprint/ai-sprint-examples

Version generated on 23/06/29

Implemented API to get the maximum admissible workload for the current
production deployment; accessing the shared volume mount point as an
environment variable. Tested on:
- blurry_faces_single_component_local_constraint
from https://gitlab.polimi.it/ai-sprint/ai-sprint-examples

Version generated on 23/06/08

Considering updated parser; updated computation of components response
times to account for the pipeline effect; using information about the
maximum admissible workload to speed-up the optimization process.
Tested on:
- mask_detection_1_partitionable_local_constraints
- mask_detection_2_partitionable_global_constraint
- mask_detection_4_double_partitioning
- mask_detection_5_three_components
- filter_classifier_degraded_performance
- filter_classifier_early_exits
from https://gitlab.polimi.it/ai-sprint/ai-sprint-examples

Version generated on 23/05/23

Relies on ai-sprint-parser version 23.05.23.

Supporting component with degraded performance and on-edge scenarios
(i.e., scenarios where only edge resources should be considered after a
network disconnection). Tested on:
- mask_detection_1_partitionable_local_constraints
- mask_detection_2_partitionable_global_constraint
- mask_detection_4_double_partitioning
- mask_detection_5_three_components
- filter_classifier_degraded_performance
from https://gitlab.polimi.it/ai-sprint/ai-sprint-examples

(note: mask_detection_4, mask_detection_5 and filter_classifier are the
only examples including edge resources)

Version generated on 23/05/03

Relies on parser version 23.05.03.

Tested on mask_detection_1_partitionable_local_constraints and
mask_detection_2_partitionable_global_constraint from
https://gitlab.polimi.it/ai-sprint/ai-sprint-examples