How Do I Load Some Well Trained Parameters During Job Training?
During job training, some parameters need to be loaded from a pre-trained model to initialize the current model. You can use the following methods to load the parameters:
- View all parameters by using the following code.
from moxing.tensorflow.utils.hyper_param_flags import mox_flags print(mox_flags.get_help())
- Specify the parameters to be restored during model loading. checkpoint_include_patterns is the parameter that needs to be restored, and checkpoint_exclude_patterns is the parameter that does not need to be restored.
checkpoint_include_patterns: Variables names patterns to include when restoring checkpoint. Such as: conv2d/weights. checkpoint_exclude_patterns: Variables names patterns to include when restoring checkpoint. Such as: conv2d/weights.
- Specify a list of parameters to be trained. trainable_include_patterns is a list of parameters that need to be trained, and trainable_exclude_patterns is a list of parameters that do not need to be trained.
--trainable_exclude_patterns: Variables names patterns to exclude for trainable variables. Such as: conv1,conv2. --trainable_include_patterns: Variables names patterns to include for trainable variables. Such as: logits.
Feedback
Was this page helpful?
Provide feedbackThank you very much for your feedback. We will continue working to improve the documentation.