-
Notifications
You must be signed in to change notification settings - Fork 227
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
LeGR implementation #501
LeGR implementation #501
Conversation
@AlexKoff88 please, take a look. |
General comments:
|
One possible scenario for parallelization is to manually wrap the model with DataParallel during LeGR search inside the algo and unwrap it after initialization so that the user can utilize whatever training mode they want later on. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
General comment - use lowercase for all .py filenames, Windows won't be able to discern capitalizations anyway
Jenkins please stop all builds |
Jenkins please retry a build |
Jenkins please stop all builds |
Jenkins please retry a build |
1 similar comment
Jenkins please retry a build |
Jenkins please stop all builds |
Jenkins please retry a build |
Jenkins please retry a build |
Jenkins please retry a build |
1 similar comment
Jenkins please retry a build |
Jenkins please stop all builds |
Ticket for documentation update: 59183 |
nncf/torch/pruning/filter_pruning/global_ranking/evolutionary_optimization.py
Show resolved
Hide resolved
Results of reproducing this algo after all merges: CIFAR-100:
Resnet-50 (CIFAR-100)
Imagenet:
|
@mkaglins , can you please update the table below by
|
Done. |
@@ -136,11 +141,50 @@ def __init__(self, target_model: NNCFNetwork, | |||
|
|||
self.weights_normalizer = tensor_l2_normalizer # for all weights in common case | |||
self.filter_importance = FILTER_IMPORTANCE_FUNCTIONS.get(params.get('weight_importance', 'L2')) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
(Issue 59470) @mkaglins, BTW, torch still expects weight_importance
attribute
To run an experiment with LeGR algorithm "compression" config section schould looks like:
Settings in the "optimizer" section will be used as settings for the final fine-tuning optimizer.