GitLab

MIGraphX verification by default uses normalized RMS error as the basis for the verification.  This change adds some logic to allow migraphx to do "np.allclose" type of elementwise verification using atol and rtol.

Commit also includes changes to consistently pass "gold" or "expected" results as the second argument for "verify_range()" calls.  Default RMS tolerance inside driver is set to 0.001 which IMO is high for FP32 compared to what we had earlier. Need better defaults

Skip verifying literals and parameters
Print the correct trim value
Print failure for an exception and continue reduce
Print if shapes do not match

MIOpen fusions are not serialized with tuned solutions. Print warnings for such cases.

Useful for CI purposes as well as keeping track in our .deb creation which
version of Onnxruntime MIGraphX has had CI tested against it.

The --py output can be loaded back in the driver. This will embed the migraphx interperter so we can execute the python directly. There is a migraphx_py library which will dynamically load the version of the library for python version is available on the system.

* Allow quantizing for both int8 and fp16

add option to print tf supported ops
sort both onnx and tf ops alphabetically

Examples..

bin/driver verify /codes/onnx_models/resnet50-v1-7/resnet50-v1-7.onnx --split-single-dyn-dim --batch 3 --dyn-input-dim @data "[{min:1, max:4}, 3, 224, 224]"

bin/driver compile /codes/onnx_models/resnet50-v1-7/resnet50-v1-7.onnx --split-single-dyn-dim --default-dyn-dim "{min:1, max:10}" --output resnet50_batch1-10.mxr

bin/driver perf resnet50_batch1-10.mxr --batch 4

prevent dynamically loading the target library that is not compiled with the same version of MIGraphX core lib.

Recent changes #1608 removed migraphx_all_target lib from driver and that led to missing compile time definitions.
Missing compile definitions led to change of default target in driver.

Fixes #1595

* Add driver flag "--exhaustive-tune" to enable tuning, add support for the same in C/C++ and python API

* Print python code

use_dynamic_same_auto_pad was removed from convolution, but the driver models still retain the fields. This PR regenerates the files so that they are compatible again.

* Enable non-standard shape
* Use perfdb for non xdlops
* Fix transpose+broadcast strides
Co-authored-by: jungpark-mlir <jungwook.park@amd.com>

Rewrites the TF batch norm like operators to other MIGX operators
Removes the code related to batch_norm_inference

* fix json strings in driver models

* Improve type printing in driver
* Improve error with incorrect order for command
* Add spell checking of arguments
* Add validations and required checking
* Add required arguments and groups

 Compiles significantly faster than constructing all the objects. It also reduces recompiles as well.

Updated each source file in the repo with the existing license.

Fixes the #error when using cppcheck. This no longer suppresses cppcheck errors when including those errors. This fixes the cppcheck errors that was there already.

* Show the operator fields in the driver

Changed the pooling values for two structures from strings to specialized enum classes. Many test and operator parsing changes to support this. Introduces one new source file, op_enums.cpp.

Allows --fp16 to be used in the driver to compare the target fp16 result and the ref fp32 result.

Currently we have the option of passing in --batch to the driver to change the batch size when the model has a dynamic dim value. We can use this flag to adjust the perf report's rate.

Added features:
This enables wrapping each migraphx operator with rocTX markers.
It adds new knob trace to migraphx-driver binary.

Limitation:

rocTX standalone does not output a file, it needs to be used with rocprof. Example command line:

/opt/rocm/bin/rocprof -i ./in.txt --hip-trace --roctx-trace --flush-rate 10ms --timestamp on -d cagri_out --obj-tracking on /opt/rocm/bin/migraphx-driver trace ./resnet50-v2-7.onnx --onnx --gpu
Co-authored-by: Shucai Xiao <shucai@gmail.com>

Previously dot operator was defined as C = alpha * A . B + beta * C where * is scalar multiplication and . is dot product or matrix multiplication depending on dimension of the inputs.

Aim is to have the definition of dot operator as C = A . B without having alpha or beta.

In order to achieve the same effect as alpha and beta (1) it multiplies the one of the inputs to the dot operator with alpha value. (2) if beta is present then, multiplies the C with beta and then adds into the output from step 1.

This reverts commit 9e43cb8b.

This PR aims to remove alpha and beta attributes from dot operator completely.

Previously dot operator was defined as C = alpha * A . B + beta * C where * is scalar multiplication and . is dot product or matrix multiplication depending on dimension of the inputs.

Aim is to have the definition of dot operator as C = A . B without having alpha or beta.

In order to achieve the same effect as alpha and beta (1) it multiplies the one of the inputs to the dot operator with alpha value. (2) if beta is present then, multiplies the C with beta and then adds into the output from step 1.

* Add a command to list supported onnx operators

* Add version command and option to driver to print migraphx version being used

* Address review comments

* Address review comments

* Fix leftover mistake

* Add newline to version.h.in

* Address review comments

* Fix tidy warning

* Remove unneccesary line from cmake file

* Formatting; remove .o files
Co-authored-by: mvermeulen <5479696+mvermeulen@users.noreply.github.com>

* fix relu6

* add more transposes

* add multi output

* formatting

* add tests

* formatting

* fix tests

* change to_nchw for outputs

* add python api

* fix cppcheck

* remove variable

* fix lambda

* add multi_output test

* add more tests and merge

* fix help message
Co-authored-by: mvermeulen <5479696+mvermeulen@users.noreply.github.com>
Co-authored-by: Paul Fultz II <pfultz2@yahoo.com>