fix pruner bugs and add model compression README #1624
Conversation
| pruner.update_epoch(epoch) | ||
|
|
||
| pruner.update_epoch(epoch + 1) |
There was a problem hiding this comment.
Let's discuss whether epoch should start from 0 or 1 next time.
| @@ -2,7 +2,7 @@ | |||
| import torch | |||
| from .compressor import Pruner | |||
|
|
|||
| __all__ = ['LevelPruner', 'AGP_Pruner', 'SensitivityPruner'] | |||
| __all__ = ['LevelPruner', 'AGP_Pruner'] | |||
There was a problem hiding this comment.
you removed sensitivitypruner, so please update doc accordingly
examples/model_compress/README.md
Outdated
| python main_torch_pruner.py | ||
| ``` | ||
|
|
||
| Model compression can be configured in 2 ways |
There was a problem hiding this comment.
This example uses AGP Pruner. Initiating a pruner needs a user provided configuration which can be provided in two ways:
examples/model_compress/README.md
Outdated
|
|
||
| Model compression can be configured in 2 ways | ||
|
|
||
| - By reading ```configure_example.yaml```, this can make codes clean when your configuration is complicated |
examples/model_compress/README.md
Outdated
| pruner = AGP_Pruner(configure_list) | ||
| ``` | ||
|
|
||
| Please notice that when ```pruner(model)``` called, our model compression codes will be **automatically injected** and you can fine-tune your model **without** any modifications, masked weights **won't** be updated any more during fine tuning. |
There was a problem hiding this comment.
When pruner(model) is called, your model is injected with masks as embedded operations. For example, a layer takes a weight as input, we will insert an operation between the weight and the layer, this operation takes the weight as input and outputs a new weight applied by the mask. Thus, the masks are applied at any time the computation goes through the operations. You can fine-tune your model without any modifications.
| @@ -8,7 +8,6 @@ We have provided two naive compression algorithms and four popular ones for user | |||
| |---|---| | |||
There was a problem hiding this comment.
“We have provided two naive compression algorithms and four popular ones for users, including three pruning algorithms and three quantization algorithms:” this line should also be updated
| @@ -48,7 +48,7 @@ from nni.compression.tensorflow import AGP_Pruner | |||
| config_list = [{ | |||
| 'initial_sparsity': 0, | |||
| 'final_sparsity': 0.8, | |||
| 'start_epoch': 1, | |||
There was a problem hiding this comment.
what is the meaning of start_epoch=0, end_epoch=10?
There was a problem hiding this comment.
@QuanluZhang start_epoch=0, end_epoch=10 means pruning starts from epoch 0 and ends at 10, previous default start_epoch=1 in the algorithm but we usually start from 0, so I modified them all.
fix pruner bugs and add model compression README