GitLab

* Add initial ck_gemm code

* Format

* Add additional src files

* Format

* Add include

* Simplify fuse_ck

* Format

* Rename var

* Enable pass

* Update ck version

* Fix include

* Add group stride

* Disable warnings for ck headers

* Format

* Add unpack array

* Add interface to enable tuning

* Format

* Update compile_ops to handle tuning config

* Format

* Add some comments

* Move time_op to migraphx_gpu

* Add banchmarking

* Refactor

* Format

* Add lift class macro

* Use device name

* Format

* Generate configs

* Format

* Pass tuning parameter

* Move data type to is_ck_gemm matcher

* Format

* Add problem_cache to avoid retuning same configs

* Format

* Format

* Mark the problems

* Format

* Use is_null

* Format

* Resize vector

* Only tune with exaustive tuning

* Format

* Use assert

* FOrmat

* Tidy fixes

* More tidy fixes

* Format

* Add license to missing files

* Format

* Use transform

* Format

* Fix tidy

* Format

* Fix cppcheck issues

* Format

* Add static_assert

* Add ops header

* Add assertion in batcher

* Format

* Improve the batch fold check

* Format

* Add where op workaround for CK

* Skip if any input is not a supported ck type

* Format

* Check batch is standard

* Format

* Remove redundant static keyword

* Update commit hash

* Fix error when running without --exhaustive-tune

* Formatting

* Formatting

* Remove fuse_ck_gemm_softmax_gemm

* Update ck hash

* Correct spelling mistake

* Remove commented out logic from fuse_ck

* Remove unused include and add comment

* Formatting

* Remove redundant get_shape and remove ck_gemm from names

* Formatting

* Allow for mixed types with int8 gemms

* Formatting

* Add back find_package from merge

* Update CK commit hash and add gfx940 to fuse_ops supported archs

* Formatting

* Update CK hash

* Fix convert for the NaNs

* NaNs can't be compared, use std::isnan()

* formatting

* formatting

* formatting

* add extra tests

* initial

* Added tests and new functionality

* Update optimals handling

* Simplify conditionals

* Ref test, update docs

* Remove comment, suggestion unclear

---------
Co-authored-by: Umang Yadav <29876643+umangyadav@users.noreply.github.com>

* fix parse_instancenorm to create broadcast and multibroadcast instructions with two dynamic shape arguments instead of 1.  Their make_op() functions don't support dynamic shapes when called with one input.  This caused an error when parsing an ONNX 3duunet model

* Use add_common_op() to create multibroadcast op.

* add verification and parsing test for instance_norm with dynamic input.  Parse test doesn't pass.

* fix for test; still doesn't pass

* another fix for test; still doesn't pass

* work in progress, instance_norm_dyn_batch_test works but instance_norm_test doesn't

* fix onnx instancenorm tests to match parser changes.  Passes all check tests

* Updated comments explaining usage of add_common_op()

* hand-merged conflicts with develop

* fix instance_norm_half_test after merge

* add Onnx test instance_norm_dyn_batch_half_test

* add shape test cases broadcast_1in_dyn_error and multibroadcast_1in_dyn_error_0

* add fix for the trace_eval

* Add throw for the debug builds

* Formatting

---------
Co-authored-by: Chris Austen <causten@users.noreply.github.com>

#1791 Added hash function for value class. It uses the Visit function and has specialization for the bool_type and <vector> type but was missing specialization for the nullptr. Nullptr caused compilation issues for RHEL, SLES and CentOS.

Enable with MIGRAPHX_ENABLE_CK=1 and --exhaustive-tune tune flag

Sigmoid approximation for GeLU was introduced in #1299 for Fp16. The sigmoid approximation is known to get better perf but lower accuracy. https://arxiv.org/pdf/1606.08415.pdf

Changed the doc for find_permutation(shape) to be more clear that it is finding the permutation that would make the shape standard

By converting to fp32 : fp16 3d-unet model accuracy comes out the same as FP32 accuracy.

By using reduce_sum method on Fp16 : accuracy comes out ~0.9% lower compared to fp32 while keeping entire model in fp16.

partially solves #1656
This PR only handles compilation part of multitarget.

Use generate_argument instead of generate_literal for python output as generate_literal doesnt exists
Shorten the names for variables from the main module
Use prefix p_ for parameters
Use shorter variable m for main module in python

* Allow quantizing for both int8 and fp16

Use std::numeric_limits::min/max() functions plus the appropriate value to encode -inf/inf 

Enable retrieving the code object to do tuning in the future.

Refactor supported gfx archs

back out changes for rocm-5.5

* use half hip functions to compute max and min
* add verify test for min and max

Co-authored-by: Paul Fultz II <pfultz2@yahoo.com>

Move CI to support the rocm5.5 release

Adding support for broadcasted scalars to unsqueeze op.

Specifying steps other than 1 is disallowed in this implementation since we want the output the always be a tensor. We can support varying step sizes if we allow a broadcasted scalar output from this op.

Remove various file not required for what we use Github runners for

Python API with documentation updates

Adds support for slice,transpose,contigous and reshape fusions into input tensors for a fused mlir kernel.

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

When multiplying either the input or output across the K dimensions then the multiple can be applied to the constant which can then be folded with propagate_const.

Exposed the mlir_enabled() call the decide for lowering pipeline's enablement
Disabled the rewrite quantization pipeline in mlir compilation
Added quant convolution as anchor ops
Fixed the return type expectations
Added the fall back hip implementation for quantizelinear and dequantizelinear
Will need advises to improve the implementation for quantizelinear

Relies on Removed split_single_dyn_dim compile flag #1711
Exposes dynamic_dimension as a opaque object with dynamic_dimensions and optimals
Exposes ONNX dyn_input_dims and default_dyn_dim to run with dynamic batch
Updates api.py to be able to create objects from aggregate initialization (used for dynamic_dimension)
Uses offload copy for now

Improves the constant propagation for bert models. Larger batch size no longer use as large of constants.  Also improves the speed of model compilation 

update rocBLAS version check to support 3.0 and above with simplified logic