Merge branch 'develop' into msgpack-large-bin-buffer

.github/workflows/sync-onnxrt-main.yaml

@@ -47,6 +47,7 @@ jobs:
             onnxruntime
             dependancies
             automated 
+            skip bot checks
           assignees: TedThemistokleous
           reviewers: TedThemistokleous causten
           draft: false

CHANGELOG.md

@@ -2,8 +2,49 @@
 
 Full documentation for MIGraphX is available at [MIGraphX Documentation](https://rocmdocs.amd.com/projects/AMDMIGraphX/en/latest/).
 
-## MIGraphX 2.5 for ROCm 5.5.0
+## MIGraphX 2.7 for ROCm 5.7.0
+### Added
+- Enabled hipRTC to not require dev packages for migraphx runtime and allow the ROCm install to be in a different directory than it was during build time
+- Add support for multi-target execution
+- Added Dynamic Batch support with C++/Python APIs
+- Add migraphx.create_argument to python API
+- Added dockerfile example for Ubuntu 22.04
+- Add TensorFlow supported ops in driver similar to exist onnx operator list
+- Add a MIGRAPHX_TRACE_MATCHES_FOR env variable to filter the matcher trace
+- Improved debugging by printing max,min,mean and stddev values for TRACE_EVAL = 2
+- use fast_math flag instead of ENV flag for GELU
+- Print message from driver if offload copy is set for compiled program
+### Optimizations
+- Optimized for ONNX Runtime 1.14.0
+- Improved compile times by only building for the GPU on the system
+- Improve performance of pointwise/reduction kernels when using NHWC layouts
+- Load specific version of the migraphx_py library
+- Annotate functions with the block size so the compiler can do a better job of optimizing 
+- Enable reshape on nonstandard shapes
+- Use half HIP APIs to compute max and min
+- Added support for broadcasted scalars to unsqueeze operator
+- Improved multiplies with dot operator
+- Handle broadcasts across dot and concat
+- Add verify namespace for better symbol resolution
+### Fixed
+- Resolved accuracy issues with FP16 resnet50
+- Update cpp generator to handle inf from  float
+- Fix assertion error during verify and make DCE work with tuples
+- Fix convert operation for NaNs
+- Fix shape typo in API test
+- Fix compile warnings for shadowing variable names
+- Add missing specialization for the `nullptr` for the hash function
+### Changed
+- Bumped version of half library to 5.6.0
+- Bumped CI to support rocm 5.6
+- Make building tests optional
+- replace np.bool with bool as per numpy request
+### Removed
+- Removed int8x4 rocBlas calls due to deprecation
+- removed std::reduce usage since not all OS' support it
+
 
+## MIGraphX 2.5 for ROCm 5.5.0
 ### Added
 - Y-Model feature to store tuning information with the optimized model
 - Added Python 3.10 bindings 
@@ -12,15 +53,11 @@ Full documentation for MIGraphX is available at [MIGraphX Documentation](https:/
 - Build support for ROCm MLIR
 - Added migraphx-driver flag to print optimizations in python (--python)
 - Added JIT implementation of the Gather and Pad operator which results in better handling of larger tensor sizes.
-
-
 ### Optimizations
 - Improved performance of Transformer based models
 - Improved performance of the Pad, Concat, Gather, and Pointwise operators
 - Improved onnx/pb file loading speed
 - Added general optimize pass which runs several passes such as simplify_reshapes/algebra and DCE in loop.
-
-
 ### Fixed
 - Improved parsing Tensorflow Protobuf files 
 - Resolved various accuracy issues with some onnx models
@@ -29,6 +66,5 @@ Full documentation for MIGraphX is available at [MIGraphX Documentation](https:/
 - Use --offload-arch instead of --cuda-gpu-arch for the HIP compiler
 - Changes inside JIT to use float accumulator for large reduce ops of half type to avoid overflow.
 - Changes inside JIT to temporarily use cosine to compute sine function.
-
 ### Changed
 - Changed version/location of 3rd party build dependencies to pick up fixes

Jenkinsfile

@@ -114,10 +114,10 @@ rocmtest clang_debug: rocmnode('cdna') { cmake_build ->
         cmake_build(flags: "-DCMAKE_BUILD_TYPE=release")
         stash includes: 'build/*.deb', name: 'migraphx-package'
     }
-}, hidden_symbols: rocmnode('cdna') { cmake_build ->
-    stage('Hidden symbols') {
-        cmake_build(flags: "-DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DCMAKE_CXX_VISIBILITY_PRESET=hidden -DCMAKE_C_VISIBILITY_PRESET=hidden")
-    }
+// }, hidden_symbols: rocmnode('cdna') { cmake_build ->
+//     stage('Hidden symbols') {
+//         cmake_build(flags: "-DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DCMAKE_CXX_VISIBILITY_PRESET=hidden -DCMAKE_C_VISIBILITY_PRESET=hidden")
+//     }
 }, all_targets_debug : rocmnode('cdna') { cmake_build ->
     stage('All targets Release') {
         cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DMIGRAPHX_ENABLE_FPGA=On") 

docs/.sphinx/requirements.in

@@ -1 +1 @@
-rocm-docs-core==0.11.0
+rocm-docs-core>=0.20.0

docs/.sphinx/requirements.txt

@@ -1,29 +1,4 @@
 # Copyright (c) 2023 Advanced Micro Devices, Inc. All rights reserved.
-
-# Permission is hereby granted, free of charge, to any person obtaining a copy
-# of this software and associated documentation files (the "Software"), to deal
-# in the Software without restriction, including without limitation the rights
-# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-# copies of the Software, and to permit persons to whom the Software is
-# furnished to do so, subject to the following conditions:
-
-# The above copyright notice and this permission notice shall be included in
-# all copies or substantial portions of the Software.
-
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
-# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
-# THE SOFTWARE.
-
-#
-# This file is autogenerated by pip-compile with Python 3.8
-# by the following command:
-#
-#    pip-compile requirements.in
-#
 accessible-pygments==0.0.4
     # via pydata-sphinx-theme
 alabaster==0.7.13
@@ -46,7 +21,7 @@ charset-normalizer==3.1.0
     # via requests
 click==8.1.3
     # via sphinx-external-toc
-cryptography==40.0.2
+cryptography==41.0.3
     # via pyjwt
 deprecated==1.2.13
     # via pygithub
@@ -60,22 +35,16 @@ fastjsonschema==2.16.3
     # via rocm-docs-core
 gitdb==4.0.10
     # via gitpython
-gitpython==3.1.31
+gitpython==3.1.32
     # via rocm-docs-core
 idna==3.4
     # via requests
 imagesize==1.4.1
     # via sphinx
-importlib-metadata==6.4.1
-    # via sphinx
-importlib-resources==5.12.0
-    # via rocm-docs-core
 jinja2==3.1.2
     # via
     #   myst-parser
     #   sphinx
-linkify-it-py==1.0.3
-    # via myst-parser
 markdown-it-py==2.2.0
     # via
     #   mdit-py-plugins
@@ -86,7 +55,7 @@ mdit-py-plugins==0.3.5
     # via myst-parser
 mdurl==0.1.2
     # via markdown-it-py
-myst-parser[linkify]==1.0.0
+myst-parser==1.0.0
     # via rocm-docs-core
 packaging==23.1
     # via
@@ -109,8 +78,6 @@ pyjwt[crypto]==2.6.0
     # via pygithub
 pynacl==1.5.0
     # via pygithub
-pytz==2023.3
-    # via babel
 pyyaml==6.0
     # via
     #   myst-parser
@@ -120,7 +87,7 @@ requests==2.28.2
     # via
     #   pygithub
     #   sphinx
-rocm-docs-core==0.11.0
+rocm-docs-core>=0.20.0
     # via -r requirements.in
 smmap==5.0.0
     # via gitdb
@@ -163,13 +130,7 @@ sphinxcontrib-serializinghtml==1.1.5
     # via sphinx
 typing-extensions==4.5.0
     # via pydata-sphinx-theme
-uc-micro-py==1.0.1
-    # via linkify-it-py
 urllib3==1.26.15
     # via requests
 wrapt==1.15.0
     # via deprecated
-zipp==3.15.0
-    # via
-    #   importlib-metadata
-    #   importlib-resources

mlir-requirements.txt

@@ -21,4 +21,4 @@
 # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 # THE SOFTWARE.
 #####################################################################################
-ROCmSoftwarePlatform/rocMLIR@ea15b3597ce55b9088621818228595dd48fb6ec0 -DBUILD_FAT_LIBROCKCOMPILER=On
+ROCmSoftwarePlatform/rocMLIR@3657f509bfed86bb79d5c6e24aa237e48f09f9f3 -DBUILD_FAT_LIBROCKCOMPILER=On

src/eliminate_contiguous.cpp

@@ -35,6 +35,8 @@
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 
+MIGRAPHX_DECLARE_ENV_VAR(MIGRAPHX_TRACE_ELIMINATE_CONTIGUOUS)
+
 static bool try_compute_shape(instruction_ref ins,
                               const std::vector<shape>& inputs,
                               const std::vector<module_ref>& mods)
@@ -78,14 +80,26 @@ static bool try_compute_shape(instruction_ref ins,
                 return (arg == ins) ? new_shape : arg->get_shape();
             });
 
-            if(not try_compute_shape(output, input_shapes, mods))
+            if(not try_compute_shape(output, input_shapes, output->module_inputs()))
             {
                 return false;
             }
         }
     }
+    catch(const std::exception& e)
+    {
+        if(enabled(MIGRAPHX_TRACE_ELIMINATE_CONTIGUOUS{}))
+        {
+            std::cout << "Exception: " << e.what() << std::endl;
+        }
+        return false;
+    }
     catch(...)
     {
+        if(enabled(MIGRAPHX_TRACE_ELIMINATE_CONTIGUOUS{}))
+        {
+            std::cout << "Unknown exception" << std::endl;
+        }
         return false;
     }
 
@@ -127,6 +141,11 @@ static void remove_contiguous(const std::string& op_name, module& m, F f)
         {
             if(arg->name() != op_name)
                 continue;
+            if(enabled(MIGRAPHX_TRACE_ELIMINATE_CONTIGUOUS{}))
+            {
+                std::cout << "eliminate_contiguous: ";
+                m.debug_print(ins);
+            }
             auto prev = arg->inputs().front();
             replace(new_args, arg, prev);
             if(try_compute_shape(ins, new_args, mod_args))

src/include/migraphx/check_shapes.hpp

@@ -1,7 +1,7 @@
 /*
  * The MIT License (MIT)
  *
- * Copyright (c) 2015-2022 Advanced Micro Devices, Inc. All rights reserved.
+ * Copyright (c) 2015-2023 Advanced Micro Devices, Inc. All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a copy
  * of this software and associated documentation files (the "Software"), to deal

src/include/migraphx/matcher.hpp

@@ -381,22 +381,24 @@ void find_matches_for(source_location location, Mod& mod, instruction_ref ins, M
     const int trace         = value_of(MIGRAPHX_TRACE_MATCHES{});
     const bool validate     = enabled(MIGRAPHX_VALIDATE_MATCHES{});
     const auto trace_filter = string_value_of(MIGRAPHX_TRACE_MATCHES_FOR{});
-    const bool trace_for    = not trace_filter.empty() and
-                           (contains(std::string{location.file_name()}, trace_filter) or
-                            contains(std::string{location.function_name()}, trace_filter));
-    bool match = false;
+    bool match              = false;
     each_args(
         [&](auto&& m) {
+            const auto& matcher_name = get_type_name(m);
+            const bool trace_for     = not trace_filter.empty() and
+                                   (contains(std::string{location.file_name()}, trace_filter) or
+                                    contains(std::string{location.function_name()}, trace_filter) or
+                                    contains(matcher_name, trace_filter));
             if(match)
                 return;
-            if(trace > 1 or trace_for)
-                std::cout << "Match: " << get_type_name(m) << std::endl;
+            if(trace > 1 and trace_for)
+                std::cout << "Match: " << matcher_name << std::endl;
             auto r = match_instruction(get_module(mod), ins, m.matcher());
             if(r.result == get_module(mod).end())
                 return;
             if(trace > 0 or trace_for)
             {
-                std::cout << "Matched by " << get_type_name(m) << std::endl;
+                std::cout << "Matched by " << matcher_name << std::endl;
                 get_module(mod).debug_print(ins);
             }
             // If its already invalid dont validate it again
@@ -407,7 +409,7 @@ void find_matches_for(source_location location, Mod& mod, instruction_ref ins, M
                 auto invalid = get_module(mod).validate();
                 if(invalid != get_module(mod).end())
                 {
-                    std::cout << "Invalid program from match: " << get_type_name(m) << std::endl;
+                    std::cout << "Invalid program from match: " << matcher_name << std::endl;
                     std::cout << "Invalid instructions: " << std::endl;
                     get_module(mod).debug_print(invalid->inputs());
                     get_module(mod).debug_print(invalid);

src/include/migraphx/normalize_attributes.hpp

@@ -1,7 +1,7 @@
 /*
  * The MIT License (MIT)
  *
- * Copyright (c) 2015-2022 Advanced Micro Devices, Inc. All rights reserved.
+ * Copyright (c) 2015-2023 Advanced Micro Devices, Inc. All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a copy
  * of this software and associated documentation files (the "Software"), to deal
@@ -28,6 +28,7 @@
 #include <migraphx/shape.hpp>
 #include <cstring>
 #include <vector>
+#include <migraphx/op/normalize_attribute.hpp>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
@@ -42,6 +43,36 @@ struct select_dependent_type
 template <class T, class... Ts>
 using dependent_type = typename select_dependent_type<T, Ts...>::type;
 
+/**
+ * Used to normalize variable input axes at model runtime.
+ * Example: the axes inputs of the slice operator.
+ *
+ * \param axes the axes to normalize
+ * \param input_shape shape of the input tensor
+ * \param attr_val the normalize_axes attributes from the operator
+ * \param prefix error message prefix
+ */
+std::vector<int64_t> normalize_axes(const std::vector<int64_t>& axes,
+                                    const shape& input_shape,
+                                    const value& attr_val,
+                                    const std::string& prefix = "");
+
+/**
+ * Used to normalize variable input axes at model runtime.
+ * Example: the starts and ends inputs of the slice operator.
+ *
+ * \param indices the indices to normalize
+ * \param axes which axes the indices apply over
+ * \param input_shape shape of the input tensor
+ * \param attr_val the normalize_axes attributes from the operator
+ * \param prefix error message prefix
+ */
+std::vector<int64_t> normalize_indices(const std::vector<int64_t>& indices,
+                                       const std::vector<int64_t>& axes,
+                                       const shape& input_shape,
+                                       const value& attr_val,
+                                       const std::string& prefix = "");
+
 MIGRAPHX_EXPORT
 bool normalize_attributes(operation& op, const shape& input_shape);
 

src/include/migraphx/op/common.hpp

@@ -1,7 +1,7 @@
 /*
  * The MIT License (MIT)
  *
- * Copyright (c) 2015-2022 Advanced Micro Devices, Inc. All rights reserved.
+ * Copyright (c) 2015-2023 Advanced Micro Devices, Inc. All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a copy
  * of this software and associated documentation files (the "Software"), to deal
@@ -33,8 +33,12 @@ namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace op {
 
+// Specifies where to add the "extra" cell of padding if the
+// calculated padding is an odd number.
 // Padding mode is default_ for fixed shape padding.
-// same_lower and same_upper used for dynamic padding.
+// same_lower and same_upper specify dynamic padding.
+// The odd cell goes at the beginning of the dimension
+// (same_lower) or end (same_upper).
 enum padding_mode_t
 {
     default_, // NOLINT