Skip to content

Conversation

@bindog
Copy link
Contributor

@bindog bindog commented Sep 25, 2019

Some of our models use the IBN-Net structure, so we need an instance norm op in TVM. I've tested on our models and it works fine.

@kevinthesun @masahi @junrushao1994 Could you have help review?

@tqchen
Copy link
Member

tqchen commented Sep 25, 2019

@hlu1 @antinucleon @vinx13 @Huyuwei can you help review the high level api ? We mainly want to make sure API consistency https://docs.tvm.ai/contribute/code_review.html#deliberate-on-api-and-data-structures

@bindog can you also comment about the related APIs in the existing packages

@bindog
Copy link
Contributor Author

bindog commented Sep 26, 2019

I update the MXNet frontend converter and test cases, I think the API is compatible with the existing packages. The instance norm is similar to relay.op.nn.nn.batch_norm except that instance norm doesn't need moving_mean and moving_var, and it also consistent with other framework like mxnet.gluon.nn. InstanceNorm

@vinx13
Copy link
Member

vinx13 commented Sep 26, 2019

I think the API is fine, as it is similar to batch_norm and layer_norm in relay.

[Relay][Op] Add instance norm op
@vinx13 vinx13 merged commit 7d911f4 into apache:master Oct 3, 2019
@vinx13
Copy link
Member

vinx13 commented Oct 3, 2019

@bindog Thanks, this is merged

anijain2305 pushed a commit to anijain2305/tvm that referenced this pull request Oct 17, 2019
* [Relay][Op] Add instance norm op

* mend

[Relay][Op] Add instance norm op
wweic pushed a commit to neo-ai/tvm that referenced this pull request Oct 18, 2019
* [Relay][Op] Add instance norm op

* mend

[Relay][Op] Add instance norm op
petrex pushed a commit to petrex/tvm that referenced this pull request Oct 29, 2019
* master: (21 commits)
  [Fix][VM] Fix VM invoke with set_params (apache#4079)
  [QNN] Refactor fixed point multiplication in requantize (apache#4073)
  Fix match case in Python-side expr functor (apache#4037)
  Hide symbols from dependent libraries if HIDE_PRIVATE_SYMBOLS is ON. (apache#4041)
  Add gradient for log-softmax (apache#4069)
  [DOC] Fix typos in tutorials (apache#4066)
  dicrease the complexity of CalcDep from exponential to linear (apache#4053)
  [Relay][AlterOp] Minor refactor. (apache#4064)
  [Relay][AlterOp] Improving support for broadcast layout alteration. (apache#4040)
  Add parses support for zeros_like tflite operator (apache#4042)
  [Bugfix][TF] reset graph after getting tag of savedmodel (apache#4055)
  [Relay][VM] Add more passes to VMCompiler (apache#4058)
  [Relay][VM] Add autotvm context when compile (apache#4062)
  [Bugfix] Fix target host for vm compiler (apache#4057)
  [Relay][Training] Add gradient for Crossentropy (apache#3925)
  [llvm] switch to use Align for llvm trunk (apache#4051)
  [Relay][TopHub] Add switch to disable TopHub download (apache#4015)
  [Relay][Op] Add instance norm op (apache#4004)
  [QNN][Relay] Calling Dialect passes from inside Relay Build API. (apache#3971)
  [RELAY/PASS] Fix the extent for the post_stmt in the loop partition (apache#3734)
  ...
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants