Add logs for user guiding better

[UsingtcNower]

Changes

Reason for changes

Related tickets

Tests

[daniil-lyakhov]

Greetings, @UsingtcNower! Thank you for your interest in the NNCF Filter Pruning algorithm!

I've checked your changes, and it looks like this warning is ambiguous: this information is present in the pruning statistics. Please check the filter pruning statistics for identity pruning masks and confirm it shows that pruning masks are identity.

In case you are using pruning algorithm in your custom pipeline, you can print statistics using following code:

compression_ctrl, compressed_model = create_compressed_model(...)
statistics = compression_ctrl.statistics()
print(statistics.to_str())

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.63%. Comparing base (6afe8d4) to head (0c3455c).
Report is 6 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff              @@
##           develop    #2523       +/-   ##
============================================
+ Coverage    46.73%   84.63%   +37.90%     
============================================
  Files          498      498               
  Lines        46178    45540      -638     
============================================
+ Hits         21579    38541    +16962     
+ Misses       24599     6999    -17600     

Files	Coverage Δ
nncf/torch/pruning/filter_pruning/algo.py	96.50% <100.00%> (+67.35%)	⬆️

... and 350 files with indirect coverage changes

Flag	Coverage Δ
COMMON	42.73% <ø> (-0.01%)	⬇️
ONNX	34.78% <0.00%> (-0.01%)	⬇️
OPENVINO	?
TENSORFLOW	29.65% <0.00%> (?)
TORCH	66.00% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
common	93.10% <ø> (+23.64%)	⬆️
torch	93.42% <100.00%> (+60.06%)	⬆️
tensorflow	93.96% <ø> (+93.96%)	⬆️
onnx	93.05% <ø> (ø)
openvino	25.58% <ø> (-68.48%)	⬇️
ptq	67.30% <ø> (-9.99%)	⬇️

[UsingtcNower]

Greetings @UsingtcNower! Thank you for your interest in the NNCF Filter Pruning algorithm!

I've checked your changes, and it looks like this warning is ambiguous: this information is present in the pruning statistics. Please check the filter pruning statistics for identity pruning masks and confirm it shows that pruning masks are identity.

Thanks for your reply @daniil-lyakhov !
Let me illustrate what I'm going through, the printed filter pruning statistics show that Pruning Level is zero, but I can't tell why. The filter pruning statistics looks like a summary, and the warning here looks like a check. Hopefully it can help users like me.

In case you are using pruning algorithm in your custom pipeline, you can print statistics using following code:

compression_ctrl, compressed_model = create_compressed_model(...)
statistics = compression_ctrl.statistics()
print(statistics.to_str())

[daniil-lyakhov]

@UsingtcNower, could you please attach NNCF logs with this warning turned on?

[UsingtcNower]

@UsingtcNower, could you please attach NNCF logs with this warning turned on?

@daniil-lyakhov , sure! Here I attach the NNCF logs.
nncf_logs.log

[alexsu52]

@UsingtcNower Thanks for the contribution!