TFLite Tutorial broken by recent commit.

[mshawcroft]

The TFLite tutorial is here: https://docs.tvm.ai/tutorials/frontend/from_tflite.html is broken by this recent commit:

#3141

commit b63267b
Author: Zhao Wu wuzhaozju@gmail.com
Date: Wed May 22 10:44:47 2019 +0800

[TFLite] Convert TFLite NCHW to NHWC (#3141)
* Convert TFLite NCHW to NHWC 
* Minor comment fix

Building the example in the tutorial results in this failure:
In main:
v0.0.1
fn (%input: Tensor[(1, 3, 224, 224), float32]) {
%0 = nn.pad(%input, pad_width=[[0, 0], [0, 1], [0, 1], [0, 0]])
%1 = nn.conv2d(%0, meta[relay.Constant][0], strides=[2, 2], channels=32, kernel_size=[3, 3], data_layout="NHWC", kernel_layout="HWIO")an internal invariant was violated while typechecking your program [15:33:02] /workspace/src/relay/pass/type_solver.cc:119: Check failed: resolved.defined(): Unable to unify parent types: TensorType([3, 3, 224, 32], float32) and TensorType([3, 3, 3, 32], float32)
;

[apivovarov]

Are you sure you are using the latest TVM version?

[mshawcroft]

Just re-checked this on master (f731e0e) and the issue is still present.

[FrozenGene]

I think you don't update the tutorial file. Your input is 1 3 224 224, which is NCHW input previously. However, current tutorial use 1 224 224 3 NHWC layout. Please checkout the latest TVM codebase of all files.

[mshawcroft]

Yep, my bad, the tutorial code I used was out of step. Thank you for looking into this, much appreciated.