Adding a new Operators: Layer Wrappers in Keras Frontend

tico · June 18, 2019, 3:19pm

Hi,

I am testing Keras models with wrappers such as TimeDistributed and Bidirectional:

However, this seems not to be supported in TVM as I get a tvm.error.OpNotImplemented. Any thoughts on how to solve this problem? I know there is a guide to add new operators, but it would be nice if someone could give me a hint on how to properly add these Keras operators.

Thanks

tico · June 18, 2019, 3:19pm

I found that TimeDistributed and Biderectional are listed but commented out in the following file in lines 595 and 596:

github.com

dmlc/tvm/blob/master/nnvm/python/nnvm/frontend/keras.py

# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
# pylint: disable=invalid-name, import-self
"""Keras frontend."""
from __future__ import absolute_import as _abs
import sys

This file has been truncated. show original

Is this support for this TBD? If so any estimation about the effort that would require to implement it?