When to support HRNet for pruning?

[ZexinYan]

Environment:

• NNI version: 2.2
• NNI mode (local|remote|pai): local
• Client OS: Linux
• Server OS (for remote mode only): Linux
• Python version: 3.8.3
• PyTorch/TensorFlow version: 1.7.1 pytroch
• Is conda/virtualenv/venv used?: coda
• Is running in Docker?: not

Log message:

• nnimanager.log:
• dispatcher.log:
• nnictl stdout and stderr:

What issue meet, what's expected?:
When to support HRNet for pruning?
How to reproduce it?:

Additional information:

[colorjam]

Hi @ZexinYan, we currently support pruning for HRNet.
Could you provide more details about what pruner you use that isn't working?

[ZexinYan]

Hi @ZexinYan, we currently support pruning for HRNet.
Could you provide more details about what pruner you use that isn't working?

Thanks for your reply. Could you provide some examples about pruning for HRNet in Cifar10? I can only find resnet or vgg as example.

[ZexinYan]

In master branch, I got the following error:

RuntimeError: Has not supported infering output shape from input shape for module/function: `prim::TupleUnpack`, .prim::TupleUnpack.201

[J-shang]

Hi, @ZexinYan
Here has a similar issue #3645 for your reference, and this issue will be fixed in pr #3462. You could have a try when this pr merged, or clone this branch then install and try.