Closed CarolinaFurtado closed 3 years ago
Resolving issue #32 - enable training starting from pretrained weights
Tested on tf2.1. Can be tested as e.g:
python3 train_segmentation_model.py --gcp-bucket gs://necstlab-sandbox --config-file configs/config_sandbox/tf_version_debug//train-small-3class_tfversion_debug.yaml --random-module-global-seed 1 --numpy-random-global-seed 1 --tf-random-global-seed 1 --pre-trained-model-id segmentation-model-small-3class_tfversion_debug_20201019T174848Z
If no --pre-trained-model-id is given, it is initiated as none and random weights are given.
--pre-trained-model-id
The metadata of the pretrained model is also included:
Resolving issue #32 - enable training starting from pretrained weights
Tested on tf2.1. Can be tested as e.g:
python3 train_segmentation_model.py --gcp-bucket gs://necstlab-sandbox --config-file configs/config_sandbox/tf_version_debug//train-small-3class_tfversion_debug.yaml --random-module-global-seed 1 --numpy-random-global-seed 1 --tf-random-global-seed 1 --pre-trained-model-id segmentation-model-small-3class_tfversion_debug_20201019T174848ZIf no
--pre-trained-model-idis given, it is initiated as none and random weights are given.The metadata of the pretrained model is also included: