Merge branch 'develop' of https://github.com/ROCmSoftwarePlatform/AMDMIGraphX into mi100_opts

.clang-tidy

@@ -5,8 +5,6 @@ CheckOptions:
     value:           '::std::async;::std::launder;::std::remove;::std::remove_if;::std::unique;::std::unique_ptr::release;::std::basic_string::empty;::std::vector::empty;::std::find;::std::find_if;::std::find_if_not;::std::all_of;::std::any_of;::std::none_of;::std::count;::std::count_if;::std::mismatch;::std::find_end;::std::find_first_of;::std::adjacent_find;::std::search;::std::search_n;::std::nth_element;::std::lower_bound;::std::upper_bound;::std::binary_search;::std::equal_range;::std::max;::std::max_element;::std::min;::std::min_element;::std::minmax;::std::minmax_element;::std::equal;::std::lexicographical_compare;::std::accumulate;::std::inner_product'
   - key:             cppcoreguidelines-macro-usage.AllowedRegexp
     value:           'DEBUG|FALLTHROUGH|STRINGIZE|_HAS_|_THROW|_REQUIRES|_DECLARE_|_VISIT_|_REGISTER_|_GENERATE_|_DETAIL_|_TIDY_|_MANAGE_PTR|_MATCHER|DEVICE_SHARED'
-  - key:             cppcoreguidelines-narrowing-conversions.WarnOnFloatingPointNarrowingConversion
-    value:           0
   - key:             modernize-loop-convert.MinConfidence
     value:           risky   
   - key:             modernize-loop-convert.NamingStyle
@@ -111,7 +109,7 @@ CheckOptions:
     value:           CamelCase
   - key:             readability-identifier-naming.TypeAliasCase
     value:           lower_case
-  # - key:             readability-identifier-naming.MacroDefinitionCase
-  #   value:           UPPER_CASE
-  # - key:             readability-identifier-naming.MacroDefinitionPrefix
-  #   value:           MIGRAPHX_
+  - key:             readability-identifier-naming.MacroDefinitionCase
+    value:           UPPER_CASE
+  - key:             readability-identifier-naming.MacroDefinitionPrefix
+    value:           MIGRAPHX_

.github/workflows/ci.yaml

@@ -151,6 +151,9 @@ jobs:
           - debug
           - release
           - codecov
+        exclude:
+          - os: ubuntu-16.04
+            configuration: debug
 
     steps:
     - uses: actions/checkout@v2

CMakeLists.txt

@@ -36,7 +36,7 @@ find_package(nlohmann_json 3.8.0 REQUIRED)
 
 include(ROCMSetupVersion)
 
-rocm_setup_version(VERSION 1.1)
+rocm_setup_version(VERSION 1.3)
 set(MIGRAPHX_SO_VERSION ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR})
 
 option( BUILD_SHARED_LIBS "Build as a shared library" ON )
@@ -148,6 +148,7 @@ rocm_enable_clang_tidy(
         -*-avoid-c-arrays
         -*-explicit-constructor
         -*-magic-numbers
+        -*-narrowing-conversions
         -*-non-private-member-variables-in-classes
         -*-use-auto
         -*-use-emplace

Dockerfile

@@ -6,7 +6,7 @@ ARG PREFIX=/usr/local
 RUN dpkg --add-architecture i386
 
 # Add rocm repository
-RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/4.1/ xenial main > /etc/apt/sources.list.d/rocm.list'
+RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/4.2/ xenial main > /etc/apt/sources.list.d/rocm.list'
 
 # Install dependencies
 RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \

cmake/PythonModules.cmake

@@ -2,27 +2,36 @@ if(COMMAND find_python)
     return()
 endif()
 
+macro(py_exec)
+    execute_process(${ARGN} RESULT_VARIABLE RESULT)
+    if(NOT RESULT EQUAL 0)
+        message(FATAL_ERROR "Process failed: ${ARGN}")
+    endif()
+endmacro()
+
 set(PYBIND11_NOPYTHON On)
 find_package(pybind11 REQUIRED)
 macro(find_python version)
     find_program(PYTHON_CONFIG_${version} python${version}-config)
     if(EXISTS ${PYTHON_CONFIG_${version}})
-        execute_process(COMMAND ${PYTHON_CONFIG_${version}} --includes OUTPUT_VARIABLE _python_include_args)
+        py_exec(COMMAND ${PYTHON_CONFIG_${version}} --includes OUTPUT_VARIABLE _python_include_args)
         separate_arguments(_python_includes UNIX_COMMAND "${_python_include_args}")
         string(REPLACE "-I" "" _python_includes "${_python_includes}")
         add_library(python${version}::headers INTERFACE IMPORTED GLOBAL)
         set_target_properties(python${version}::headers PROPERTIES
             INTERFACE_INCLUDE_DIRECTORIES "${_python_includes}"
         )
-        execute_process(COMMAND ${PYTHON_CONFIG_${version}} --prefix OUTPUT_VARIABLE _python_prefix)
+        py_exec(COMMAND ${PYTHON_CONFIG_${version}} --prefix OUTPUT_VARIABLE _python_prefix)
         string(STRIP "${_python_prefix}" _python_prefix)
         set(PYTHON_${version}_EXECUTABLE "${_python_prefix}/bin/python${version}" CACHE PATH "")
     endif()
 endmacro()
 function(py_extension name version)
-    set(_python_module_extension)
-    execute_process(COMMAND ${PYTHON_CONFIG_${version}} --extension-suffix OUTPUT_VARIABLE _python_module_extension)
+    set(_python_module_extension ".so")
+    if(version VERSION_GREATER_EQUAL 3.0)
+        py_exec(COMMAND ${PYTHON_CONFIG_${version}} --extension-suffix OUTPUT_VARIABLE _python_module_extension)
         string(STRIP "${_python_module_extension}" _python_module_extension)
+    endif()
     set_target_properties(${name} PROPERTIES PREFIX "" SUFFIX "${_python_module_extension}")
 endfunction()
 function(py_add_module NAME)

examples/README.md

@@ -13,3 +13,5 @@ This directory contains examples of common use cases for MIGraphX.
 - [Python Resnet50 Inference](./python_api_inference)
 - [Python BERT SQuAD Inference](./python_bert_squad_example)
 - [Python Super Resolution](./python_super_resolution)
+- [Python NFNet Inference](./python_nfnet_inference)
+

examples/cpp_api_inference/README.md

@@ -118,7 +118,7 @@ This directory contains everything that is needed to perform inference on an MNI
 ```
 $ mkdir build
 $ cd build
-$ cmake ..
+$ CXX=/opt/rocm/llvm/bin/clang++ cmake ..
 $ make
 ```
 There will now be an executable named `mnist_inference` in the `build` directory. This can be run with or without options. Executing without any options will produce the following output:

examples/cpp_api_inference/mnist_inference.cpp

@@ -122,10 +122,8 @@ int main(int argc, char** argv)
 
     migraphx::program_parameters prog_params;
     auto param_shapes = prog.get_parameter_shapes();
-    for(auto&& name : param_shapes.names())
-    {
-        prog_params.add(name, migraphx::argument(param_shapes[name], digit.data()));
-    }
+    auto input        = param_shapes.names().front();
+    prog_params.add(input, migraphx::argument(param_shapes[input], digit.data()));
 
     std::cout << "Model evaluating input..." << std::endl;
     auto start   = std::chrono::high_resolution_clock::now();

examples/python_nfnet_inference/README.md

+# NFNet Inference with MIGraphX
+
+## NFNet
+NFNet: Normalizer-Free Nets. An image recognition model that can be trained without batch normalization layers. It instead uses gradient clipping algorithm to provide same affects of BatchNorm.
+
+<ins>**Summary:**</ins>
+- SOTA on ImageNet (86.5% top-1 w/o extra data)
+- Up to 8.7x faster to train than EfficientNets to a given accuracy
+- Normalizer-free (no BatchNorm)
+
+**Paper**: https://arxiv.org/pdf/2102.06171.pdf
+
+**Colab notebook**: https://github.com/deepmind/deepmind-research/tree/master/nfnets
+
+### Why not batch norm?
+
+Batch normalization has three significant practical disadvantages:
+1. It is an expensive computational primitive, which incurs memory overhead and significantly increases the time required to evaluate the gradient in some networks.
+2. It introduces a discrepancy between the behavior of the model during training and at inference time, introducing hidden hyper-parameters that have to be tuned.
+3. Last and most important point, batch normalization breaks the independence between training examples in the minibatch (batch size matters with batch norm, distributed training becomes extremely cumbersome).
+
+Instead:
+
+- Authors provide Adaptive Gradient Clipping (AGC), which clips gradients based on the unit-wise ratio of gradient norms to parameter norms, and they demonstrate that AGC allows them to train normalizer-free networks with larger batch sizes and stronger data augmentations.
+- They design a family of Normalizer-Free ResNets, called NFNets, which set new state-of-the-art validation accuracies on ImageNet for a range of training latencies. Their NFNet-F1 model achieves similar accuracy to EfficientNet-B7 while being 8.7× faster to train, and their largest model sets a new overall state of the art without extra data of 86.5% top-1 accuracy.
+- They show that NFNets achieve substantially higher validation accuracies than batch-normalized networks when fine-tuning on ImageNet after pre-training on a large private dataset of 300 million labelled images. Their best model achieves 89.2% top-1 accuracy after fine-tuning.
+
+## Inference with MIGraphX using NFNet ONNX Model
+
+There is no ONNX model released for NFNet, as of June 2021, however a PyTorch model is available at:
+https://github.com/rwightman/pytorch-image-models. 
+We provide an in-house produced and optimized ONNX model, which can be parsed and compiled using MIGraphX for AMD GPUs. The ONNX model file can be fetched using the Jupyter notebook we provide.
+
+### Requirements:
+1) AMD GPU system with ROCm installed.
+2) Jupyter notebook library.
+
+### How to use NFNet for image recognition:
+Please utilize the notebook example provided:
+1) Install jupyter notebook to your environment if not already installed:
+```
+https://jupyter.org/install
+```
+2) Connect to your jupyter server and utilize `nfnet_inference.ipynb` notebook file.
+
+### How to compare MIGraphX to ONNX Runtime for NFNet ONNX model:
+First install requirements:
+```
+pip3 install -r requirements_nfnet.txt
+```
+
+On your terminal, invoke:
+```
+python3 ort_comparison.py
+````

examples/python_nfnet_inference/nfnet_inference.ipynb

+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# NFNet Inference with AMD MIGraphX\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Normalizer-Free ResNet is a new residual convolutional network providing new state-of-the-art Top-1 accuracy of 86.5% at ImageNet dataset. The most important feature of the model is removing batch normalization. Instead of batch normalization, it uses adaptive gradient clipping to provide same regularization effect of BatchNorm. <br> Details of this network: https://arxiv.org/abs/2102.06171\n",
+    "\n",
+    "In this notebook, we are showing: <br>\n",
+    "- How to optimize NFNet ONNX model with AMD MIGraphX.\n",
+    "- How to run inference on AMD GPU with the optimized ONNX model.\n",
+    "\n",
+    "The NFNet utilized in this example is the smallest NFNet version, F0: 71.5M parameters (83.6% top-1 accuracy on ImageNet)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Requirements"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!apt-get update\n",
+    "!apt-get install ffmpeg libsm6 libxext6  -y "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!pip3 install --upgrade pip\n",
+    "!pip3 install -r requirements_nfnet.txt"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "import cv2\n",
+    "import json\n",
+    "from PIL import Image\n",
+    "import time\n",
+    "from os import path "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Importing AMD MIGraphX Python Module"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import migraphx"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Create NFNet ONNX file\n",
+    "Following repository provides functionality to create NFNet ONNX file from PyTorch model."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!wget -nc https://www.dropbox.com/s/u4ga8zyxtppfzxc/dm_nfnet_f0.onnx"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Load ImageNet labels"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "with open('../python_api_inference/imagenet_simple_labels.json') as json_data:\n",
+    "    labels = json.load(json_data)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "## Load ONNX model using MIGraphX"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model = migraphx.parse_onnx(\"dm_nfnet_f0.onnx\")\n",
+    "model.compile(migraphx.get_target(\"gpu\"))\n",
+    "\n",
+    "print(model.get_parameter_names())\n",
+    "print(model.get_parameter_shapes())\n",
+    "print(model.get_output_shapes())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Functions for image processing"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def make_nxn(image, n):\n",
+    "    height, width = image.shape[:2]    \n",
+    "    if height > width:\n",
+    "        dif = height - width\n",
+    "        bar = dif // 2 \n",
+    "        square = image[(bar + (dif % 2)):(height - bar),:]\n",
+    "        return cv2.resize(square, (n, n))\n",
+    "    elif width > height:\n",
+    "        dif = width - height\n",
+    "        bar = dif // 2\n",
+    "        square = image[:,(bar + (dif % 2)):(width - bar)]\n",
+    "        return cv2.resize(square, (n, n))\n",
+    "    else:\n",
+    "        return cv2.resize(image, (n, n))\n",
+    "    \n",
+    "def preprocess(img_data):\n",
+    "    mean_vec = np.array([0.485, 0.456, 0.406])\n",
+    "    stddev_vec = np.array([0.229, 0.224, 0.225])\n",
+    "    norm_img_data = np.zeros(img_data.shape).astype('float32')\n",
+    "    for i in range(img_data.shape[0]):  \n",
+    "        norm_img_data[i,:,:] = (img_data[i,:,:]/255 - mean_vec[i]) / stddev_vec[i]\n",
+    "    return norm_img_data\n",
+    "\n",
+    "def input_process(frame, dim):\n",
+    "    # Crop and resize original image\n",
+    "    cropped = make_nxn(frame, dim)\n",
+    "    # Convert from HWC to CHW\n",
+    "    chw = cropped.transpose(2,0,1)\n",
+    "    # Apply normalization\n",
+    "    pp = preprocess(chw)\n",
+    "    # Add singleton dimension (CHW to NCHW)\n",
+    "    data = np.expand_dims(pp.astype('float32'),0)\n",
+    "    return data"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Download example image"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Fetch example image: traffic light\n",
+    "!wget -nc http://farm5.static.flickr.com/4072/4462811418_8bc2bd42ca_z_d.jpg -O traffic_light.jpg\n",
+    "# Read the image\n",
+    "im = cv2.imread('traffic_light.jpg')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Process the read image to conform input requirements\n",
+    "data_input = input_process(im, 192)\n",
+    "\n",
+    "# Run the model\n",
+    "start = time.time()\n",
+    "results = model.run({'inputs':data_input}) # Your first inference would take longer than the following ones.\n",
+    "print(f\"Time inference took: {100*(time.time() - start):.2f}ms\")\n",
+    "# Extract the index of the top prediction\n",
+    "res_npa = np.array(results[0])\n",
+    "print(f\"\\nResult: {labels[np.argmax(res_npa)]}\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Run the model again, first one would take long\n",
+    "start = time.time()\n",
+    "results = model.run({'inputs':data_input}) # Your first inference would take longer than the following ones.\n",
+    "print(f\"Time inference took: {100*(time.time() - start):.2f}ms\")\n",
+    "# Extract the index of the top prediction\n",
+    "res_npa = np.array(results[0])\n",
+    "print(f\"\\nResult: {labels[np.argmax(res_npa)]}\")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

examples/python_nfnet_inference/ort_comparison.py

+import numpy
+import onnxruntime as rt
+
+sess = rt.InferenceSession("dm_nfnet_f0.onnx")
+
+input_name = sess.get_inputs()[0].name
+print("input name", input_name)
+input_shape = sess.get_inputs()[0].shape
+print("input shape", input_shape)
+input_type = sess.get_inputs()[0].type
+print("input type", input_type)
+
+output_name = sess.get_outputs()[0].name
+print("output name", output_name)
+output_shape = sess.get_outputs()[0].shape
+print("output shape", output_shape)
+output_type = sess.get_outputs()[0].type
+print("output type", output_type)
+
+x = numpy.random.random((1, 3, 192, 192))
+x = x.astype(numpy.float32)
+
+import migraphx
+model = migraphx.parse_onnx("dm_nfnet_f0.onnx")
+model.compile(migraphx.get_target("gpu"))
+print(model.get_parameter_names())
+print(model.get_parameter_shapes())
+print(model.get_output_shapes())
+
+result_migraphx = model.run({"inputs": x})
+result_ort = sess.run([output_name], {input_name: x})
+
+result_migraphx = result_migraphx[0].tolist()
+
+for i in range(10):
+    x = numpy.random.random((1, 3, 192, 192))
+    x = x.astype(numpy.float32)
+
+    result_migraphx = model.run({"inputs": x})
+    result_ort = sess.run([output_name], {input_name: x})
+
+    try:
+        numpy.testing.assert_allclose(result_migraphx[0].tolist(),
+                                      result_ort[0][0],
+                                      rtol=1e-02)
+        print(f"Test #{i} completed.")
+    except AssertionError as e:
+        print(e)

examples/python_nfnet_inference/requirements_nfnet.txt

+opencv-python
+onnxruntime
\ No newline at end of file

hip-clang.docker

@@ -6,7 +6,7 @@ ARG PREFIX=/usr/local
 RUN dpkg --add-architecture i386
 
 # Add rocm repository
-RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/4.1/ xenial main > /etc/apt/sources.list.d/rocm.list'
+RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/4.2/ xenial main > /etc/apt/sources.list.d/rocm.list'
 
 # Install dependencies
 RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \

src/CMakeLists.txt

@@ -23,12 +23,14 @@ add_library(migraphx
     eliminate_data_type.cpp
     eliminate_identity.cpp
     eliminate_pad.cpp
+    insert_pad.cpp
     file_buffer.cpp
     rewrite_batchnorm.cpp
     rewrite_rnn.cpp
     rewrite_pooling.cpp
     env.cpp
     generate.cpp
+    inline_module.cpp
     instruction.cpp
     load_save.cpp
     make_op.cpp
@@ -99,6 +101,7 @@ register_migraphx_ops(
     flatten
     floor
     gather
+    get_tuple_elem
     greater
     gru
     identity
@@ -135,6 +138,7 @@ register_migraphx_ops(
     reduce_sum
     relu
     reshape
+    reverse
     rnn
     rnn_last_cell_output
     rnn_last_hs_output

src/compile_src.cpp

@@ -32,7 +32,7 @@ std::vector<char> src_compiler::compile(const std::vector<src_file>& srcs) const
         }
     }
 
-    params += " -o" + out;
+    params += " -o " + out;
 
     td.execute(compiler, params);
 

src/eliminate_pad.cpp

@@ -5,76 +5,86 @@
 #include <migraphx/op/im2col.hpp>
 #include <migraphx/op/pooling.hpp>
 #include <migraphx/op/pad.hpp>
+#include <migraphx/make_op.hpp>
 #include <migraphx/iterator_for.hpp>
 #include <migraphx/stringutils.hpp>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 
-void eliminate_pad::apply(module& p) const
-{
-    for(auto ins : iterator_for(p))
-    {
-        const std::string& op_name = ins->name();
-        if(op_name != "convolution" and op_name != "im2col" and op_name != "pooling")
-            continue;
-        auto input = ins->inputs().front();
-        if(input->name() != "pad")
-            continue;
-        if(op_name == "convolution" or op_name == "im2col")
-            update_op(input, ins, p);
-        else if(op_name == "pooling")
-            update_pooling(input, ins, p);
-    }
-}
-
-void eliminate_pad::update_op(const instruction_ref& input,
-                              const instruction_ref& ins,
-                              module& p) const
+static void update_op(const instruction_ref& input, const instruction_ref& ins, module& m)
 {
     auto pad_op = any_cast<op::pad>(input->get_operator());
-    if(!pad_op.symmetric())
-        return;
 
     auto kdims    = input->get_shape().lens().size() - 2;
     auto kdims_it = pad_op.pads.begin() + 2;
 
-    std::vector<size_t> new_pads(kdims_it, kdims_it + kdims);
+    std::vector<size_t> pads_l(kdims_it, kdims_it + kdims);
+    std::vector<size_t> pads_r(kdims_it + kdims + 2, pad_op.pads.end());
 
     auto op = ins->get_operator();
-    op.from_value({{"padding", new_pads}});
+    std::vector<size_t> padding(kdims * 2, 0);
+
+    std::transform(
+        pads_l.begin(), pads_l.end(), padding.begin(), padding.begin(), std::plus<size_t>());
+    std::transform(pads_r.begin(),
+                   pads_r.end(),
+                   padding.begin() + kdims,
+                   padding.begin() + kdims,
+                   std::plus<size_t>());
+
+    op.from_value({{"padding", padding}});
 
     std::vector<instruction_ref> new_inputs{ins->inputs()};
     new_inputs.front() = input->inputs().front();
 
-    p.replace_instruction(ins, op, new_inputs);
+    m.replace_instruction(ins, op, new_inputs);
 }
 
-void eliminate_pad::update_pooling(const instruction_ref& input,
-                                   const instruction_ref& ins,
-                                   module& p) const
+static void update_pooling(const instruction_ref& input, const instruction_ref& ins, module& m)
 {
-    auto pad_op = any_cast<op::pad>(input->get_operator());
-    if(!pad_op.symmetric())
-        return;
-
-    auto kdims    = input->get_shape().lens().size() - 2;
-    auto kdims_it = pad_op.pads.begin() + 2;
-
-    std::vector<size_t> new_pads(kdims_it, kdims_it + kdims);
-
     auto op = any_cast<op::pooling>(ins->get_operator());
     if(op.mode == "average")
     {
         return;
     }
+    auto pad_op = any_cast<op::pad>(input->get_operator());
+
+    auto kdims    = input->get_shape().lens().size() - 2;
+    auto kdims_it = pad_op.pads.begin() + 2;
+
+    std::vector<size_t> pads_l(kdims_it, kdims_it + kdims);
+    std::vector<size_t> pads_r(kdims_it + kdims + 2, pad_op.pads.end());
 
-    op.padding = new_pads;
+    std::transform(
+        pads_l.begin(), pads_l.end(), op.padding.begin(), op.padding.begin(), std::plus<size_t>());
+    std::transform(pads_r.begin(),
+                   pads_r.end(),
+                   op.padding.begin() + kdims,
+                   op.padding.begin() + kdims,
+                   std::plus<size_t>());
 
     std::vector<instruction_ref> new_inputs{ins->inputs()};
     new_inputs.front() = input->inputs().front();
 
-    p.replace_instruction(ins, op, new_inputs);
+    m.replace_instruction(ins, op, new_inputs);
+}
+
+void eliminate_pad::apply(module& m) const
+{
+    for(auto ins : iterator_for(m))
+    {
+        const std::string& op_name = ins->name();
+        if(op_name != "convolution" and op_name != "im2col" and op_name != "pooling")
+            continue;
+        auto input = ins->inputs().front();
+        if(input->name() != "pad")
+            continue;
+        if(op_name == "convolution" or op_name == "im2col")
+            update_op(input, ins, m);
+        else if(op_name == "pooling")
+            update_pooling(input, ins, m);
+    }
 }
 
 } // namespace MIGRAPHX_INLINE_NS

src/include/migraphx/check_shapes.hpp

@@ -158,6 +158,13 @@ struct check_shapes
         return *this;
     }
 
+    const check_shapes& tuple_type() const
+    {
+        if(!this->all_of([](const shape& s) { return s.type() == shape::tuple_type; }))
+            MIGRAPHX_THROW(prefix() + "Shapes are not tuple!");
+        return *this;
+    }
+
     const check_shapes& not_transposed() const
     {
         if(!this->all_of([](const shape& s) { return not s.transposed(); }))

src/include/migraphx/eliminate_pad.hpp

@@ -20,10 +20,7 @@ struct eliminate_pad
 {
     std::string name() const { return "eliminate_pad"; }
 
-    void apply(module& p) const;
-    void update_op(const instruction_ref& input, const instruction_ref& ins, module& p) const;
-
-    void update_pooling(const instruction_ref& input, const instruction_ref& ins, module& p) const;
+    void apply(module& m) const;
 };
 
 } // namespace MIGRAPHX_INLINE_NS