tensorflow::ops::AddN is not supported

There is AddN operator in ResNet-50,but it is not supported by NNVM. The node is like below:
image

I got an error like that:
NotImplementedError: The following operators are not implemented: {‘AddN’}

Thanks for reporting this.

Has this issue be solved?

Not done yet, will do by this week end.

I meet it too. The file tvm/nnvm/python/nnvm/frontend/tensorflow.py ,this is not include ‘AddN’:
_convert_map = {
‘ArgMax’ : _argx(_sym.argmax, ‘argmax’),
‘ArgMin’ : _argx(_sym.argmin, ‘argmin’),
‘AvgPool’ : _pooling(‘avg_pool’),
‘BatchNormWithGlobalNormalization’ : _batch_norm(),
‘BiasAdd’ : _bias_add(),
‘Cast’ : _cast(),
‘CheckNumerics’ : _check_numerics(),
‘Concat’ : _concat(),
‘ConcatV2’ : _concatV2(),
‘Conv2D’ : _conv(‘conv’),
‘DecodeJpeg’ : _decode_image(),
‘ExpandDims’ : _expand_dims(),
‘Identity’ : _identity(),
‘MatMul’ : _matmul(),
‘MaxPool’ : _pooling(‘max_pool’),
‘Add’ : _elemwise(‘add’),
‘Sub’ : _elemwise(‘sub’),
‘Mul’ : _elemwise(‘mul’),
‘Maximum’ : _elemwise(‘max’),
‘Minimum’ : _elemwise(‘min’),
‘Sum’ : _sum(),
‘Square’ : _square(),
‘Relu’ : AttrCvt(‘relu’),
‘Reshape’ : _reshape(),
‘ResizeBilinear’ : _resize_bilinear(),
‘Softmax’ : AttrCvt(‘softmax’, {‘axis’: (‘axis’, 1)}),
‘Rsqrt’ : _rsqrt(),
‘Squeeze’ : _squeeze(),
‘FusedBatchNorm’ : _fused_batch_norm(),
‘Relu6’ : _relu6(),
‘DepthwiseConv2dNative’ : _conv(‘depthwise’),
‘Shape’ : _shape(),
‘Sigmoid’ : AttrCvt(‘sigmoid’),
‘Fill’ : _fill(),
‘GatherV2’ : _gather_v2(),
‘StridedSlice’ : _stridedSlice(),
‘LRN’ : _lrn(),
‘Pad’ : _pad(‘Pad’),
‘PadV2’ : _pad(‘PadV2’),
}

@caoqichun see python/tvm/relay/frontend/tensorflow.py

Perhaps you are using v0.5 code? nnvm is replaced by relay

I use v0.8 code, It is lastest?

This op is supported in the Relay TensorFlow frontend. AFAIK, NNVM is deprecated and you should switch to Relay.