[GitHub] [singa] dcslin commented on pull request #697: New Model Layer Operator API

[GitBox <gi...@apache.org>]

dcslin commented on pull request #697:
URL: https://github.com/apache/singa/pull/697#issuecomment-630741700


   Hi @nudles @nudles get/set params are implemented as required:
   
   ```
   {'l1:W': [[-0.838156    0.58935565]
    [ 0.4453408   0.08828315]
    [ 0.10353903  0.974195  ]
    [ 0.24744798  0.26123875]], 'l1:b': [0. 0.], 'bn1:scale': [1. 1.], 'bn1:bias': [0. 0.], 'dl1:l1:W': [[ 0.2805689  -0.05141211 -0.27061072  0.02269484]
    [-0.5174572  -0.02928859 -0.10004351 -0.47666895]], 'dl1:l1:b': [0. 0. 0. 0.], 'dl1:l2:W': [[ 1.0805224   0.01622006]
    [ 0.26961604  0.61809814]
    [ 1.0673248  -0.26117754]
    [-0.50873727  0.09084714]], 'dl1:l2:b': [0. 0.]}
   ```
   
   May I know for `states`, what is the rationale to make it separated from `params`? could we just make it as params? for now only `batchnorm` has states.
   Could we have it like together:
   ```
    'bn1:scale': [1. 1.], 'bn1:bias': [0. 0.], 'bn1:running_mean': [1. 1.], 'bn1:running_var': [0. 0.],
   ```


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org