Refactor doc of NNI model compression #2595
Conversation
QuanluZhang
requested review from
chicm-ms,
scarlett2018,
squirrelsc and
colorjam
docs/en_US/Compressor/Pruner.md
Outdated
| @@ -1,5 +1,4 @@ | |||
| Pruner on NNI Compressor | |||
| === | |||
| # Supported Pruning Algorithms on NNI | |||
|
|
|||
| Index of supported pruning algorithms | |||
There was a problem hiding this comment.
Suggest to organize with fine-grained pruning / structural pruning
There was a problem hiding this comment.
could you give me a version of such organization?
There was a problem hiding this comment.
There was a problem hiding this comment.
The WeightRankFilterPruner, ActivationRankFilterPruner and GradientRankFilterPruner have been removed from code. Suggest to flatten the structured pruners.
There was a problem hiding this comment.
And suggest move lottery ticket to another category , or same category as AGP
There was a problem hiding this comment.
updated, could @chicm-ms make further changes in your pr? thanks!
|
|
||
| The experiments code can be found at [examples/model_compress]( https://github.com/microsoft/nni/tree/master/examples/model_compress/) | ||
|
|
||
| *** | ||
|
|
||
| ## WeightRankFilterPruner |
There was a problem hiding this comment.
Would you remove the WeightRankFilterPruner? It has been removed in code.
| @@ -355,6 +387,7 @@ You can view example for more information | |||
| - **sparsity:** How much percentage of convolutional filters are to be pruned. | |||
| - **op_types:** Only Conv2d is supported in ActivationMeanRankFilterPruner. | |||
|
|
|||
| *** | |||
|
|
|||
| ## GradientRankFilterPruner | |||
There was a problem hiding this comment.
GradientRankFilterPruner can also be removed.
| @@ -292,6 +322,8 @@ pruner.compress() | |||
| - **sparsity:** This is to specify the sparsity operations to be compressed to | |||
| - **op_types:** Only Conv1d and Conv2d is supported in L2Filter Pruner | |||
|
|
|||
| *** | |||
|
|
|||
| ## ActivationRankFilterPruner | |||
There was a problem hiding this comment.
ActivationRankFilterPruner also needs to be removed.
|
|
||
| To better demonstrate how to customize a new pruning algorithm, it is necessary for users to first understand the framework for supporting various pruning algorithms in NNI. | ||
|
|
||
| ### Framework overview for pruning algorithms |
There was a problem hiding this comment.
The title doesn't help to understand purpose of its content. The content should like a step by step guidance to create user's puring algorithm.
There was a problem hiding this comment.
also leave it to @chicm-ms :)
|
@scarlett2018 @squirrelsc @colorjam I will create a new PR to address your comments. |
No description provided.