Flatten kernel refactor #8
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
* Optimization of pool2d grad, first commit. * remove useless print codes * refine codes * refine codes * seal more operation into template specialization * fix template struct error in MaxPool2dGrad. * Fix header including error * refine code with comment * Seal the param-preparation codes into function for common use. * Seal the param-preparation codes into function for common use. * Seal the param-preparation into funciton and make it common for other kernels * polish code and erase useless template speicalization * Rerun triger * rerun trigger
…addlePaddle#35510) * Create stateful OneDNNAXPYHandler object. This makes it possible to call it multiple times without recreating the oneDNN primitives every time. * Prepare SGDOpKernel to reuse its implementation from OneDNN kernel. * OneDNN SGD kernel. * Update call to use new OneDNNAXPYHandler object api. * Setup seed in proper place. * Enable OneDNN kernel only for single case. * For dense param and sparse grad. * Small refactor. * Enable oneDNN by op attr or by cmd line flag. * Use int64_t type for number of elements. * Support dense param and grad from OneDNN kernel. * Enable SGD OneDNN kernel when use MP BF16 optimizer. * Force non-copyable/movable OneDNNAXPYHandler. * Reuse OneDNNAXPYHandler for spare tensors in SUM op. * Fix SFINAE rules. * Remove recording event inside AXPY. * Get rid of internal primitive caching. * Stop use PP cache mechanims to store mem and primitive obj. * Handler obj store and reuse needed desc & prim * Do not derive from MKLDNNHandlerT
…e/Paddle into kernel_refactor_demo
chenwhql
pushed a commit
that referenced
this pull request
Jan 19, 2022
…addlePaddle#38914) * Merged LoDTensor with Tensor,test=allcases * Patched python level LoDTensor * Patched python level LoDTensor * Merge Tensor into DenseTensor * Fixed namespace issues,test=allcases * Fixed merge issues * Fixed inference issues * Fixed NPU test issues * Fixed merge issues
chenwhql
pushed a commit
that referenced
this pull request
Apr 6, 2022
…ePaddle#41387) * [Refactor] refactored eager_gen.py PR #2 * [DoubleGrad PR #1] Decoupled code generation logics for Dygraph ForwardFunctions and GradNodes * Fixed minor issue * Adjusted logics of GenerateNodeCreationCodes and GenerateForwardDefinition * Fixed issues * Supported higher-order grad node generation * [DoubleGrad PR #4] Supported higher-order GradNode generation * [DoubleGrad #4] Bug Fixes to Double Grad Node Generation * Fixed yaml typo * Fixed yaml typo * fixed minor issues * [DoubleGrad PR #5] Enabled gradient computations for grad_tensors passed to paddle.grad() * Fixed minor issue * Fixed CI-Inference issue * Fixed CI-inference issues * [DoubleGrad PR #7] paddle.grad() to copy backward graph before backward run * Fixed minor issues * Fixed issue with backward graph construction logic * Fixed implementation issues with backward graph reconstruction * Fixed unittest issue * Fixed issues * [DoubleGrad PR #8] Enabled triple grads for sigmoid and matmul * Fixed issues with phi kernel * Added triple grad test case * Fixed minor issue
chenwhql
pushed a commit
that referenced
this pull request
Mar 16, 2023
* [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * Pr 50885 (#7) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * fix code in a dy2static-friendly way. * [dystatic] add hooker for prim --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [prim] enable dygraph_to_static to support custom_vjp * fix cast prim and vjp dtype mapping error bug * [dy2static-ci] fix dy2static ci errors. --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com>
chenwhql
pushed a commit
that referenced
this pull request
Mar 16, 2023
* [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes * add unittest * fix typo * fix typo * fix map.at * fix find * fix test * fix cinn cache key structure realize * using ordered map for attributes * add test by review advice --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * Pr 50885 (#7) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes * add unittest * fix typo * fix typo * fix map.at * fix find * fix test * fix cinn cache key structure realize * using ordered map for attributes * add test by review advice --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * fix code in a dy2static-friendly way. * [dystatic] add hooker for prim --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [prim] enable dygraph_to_static to support custom_vjp * fix cast prim and vjp dtype mapping error bug * Cxx prim custom vjp (#8) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * Pr 50885 (#7) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * fix code in a dy2static-friendly way. * [dystatic] add hooker for prim --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [prim] enable dygraph_to_static to support custom_vjp * fix cast prim and vjp dtype mapping error bug * [dy2static-ci] fix dy2static ci errors. --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [Prim] enable whitelist and blacklist for custom_vjp * support softmax grad * remove additional code * add test back --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> Co-authored-by: xiongkun <807377414@qq.com>
chenwhql
pushed a commit
that referenced
this pull request
Mar 30, 2023
* Add flatten composite rule * get the right xshape and pass func test * add cinn unit test * Remove cinn test, wait for it to be added after repair * add comp test to test_flatten_contiguous_range_op.py * remove func test on composite_ops * Add comments to maybe_wrap_dim func * remove commented code * fix the problem with 0D tensor case * add flatten split rule comment * fix syntax issues * block flatten on resnet_prim_cinn * init change * tmp commit * add layer_norm InferMeta check * cast type modify * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes * add unittest * fix typo * fix typo * fix map.at * fix find * fix test * fix cinn cache key structure realize * using ordered map for attributes * add test by review advice --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * Pr 50885 (#7) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes * add unittest * fix typo * fix typo * fix map.at * fix find * fix test * fix cinn cache key structure realize * using ordered map for attributes * add test by review advice --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * fix code in a dy2static-friendly way. * [dystatic] add hooker for prim --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [prim] enable dygraph_to_static to support custom_vjp * fix cast prim and vjp dtype mapping error bug * recover * big tol * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes * add unittest * fix typo * fix typo * fix map.at * fix find * fix test * fix cinn cache key structure realize * using ordered map for attributes * add test by review advice --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * Pr 50885 (#7) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes * add unittest * fix typo * fix typo * fix map.at * fix find * fix test * fix cinn cache key structure realize * using ordered map for attributes * add test by review advice --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * fix code in a dy2static-friendly way. * [dystatic] add hooker for prim --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [prim] enable dygraph_to_static to support custom_vjp * fix cast prim and vjp dtype mapping error bug * Cxx prim custom vjp (#8) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * Pr 50885 (#7) * [CINN]Enhance CacheKey hash logic by considering input dtypes (PaddlePaddle#50557) * [CINN]Enhance CacheKey hash logic by considering input dtypes --------- Co-authored-by: jiangcheng <thisjiang@qq.com> * [prim] enable dygraph_to_static to support custom_vjp * fix code in a dy2static-friendly way. * [dystatic] add hooker for prim --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [prim] enable dygraph_to_static to support custom_vjp * fix cast prim and vjp dtype mapping error bug * [dy2static-ci] fix dy2static ci errors. --------- Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> * [Prim] enable whitelist and blacklist for custom_vjp * debug log * clear log * fix * nothing * less memory * recover utils * fix * modify threshold value * skip layer_norm for test_bert * back to bert success state * add epsion * delete unnecessary compute * modify amp dtype * modify * order * delete sqrt check and fp16 --------- Co-authored-by: xuyongsheng <xuyongsheng@baidu.com> Co-authored-by: xysheng-baidu <121540080+xysheng-baidu@users.noreply.github.com> Co-authored-by: Aurelius84 <zhangliujie@baidu.com> Co-authored-by: jiangcheng <thisjiang@qq.com> Co-authored-by: cxxly <chenxx_id@163.com> Co-authored-by: xiongkun <807377414@qq.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR types
Others
PR changes
OPs
Describe
flatten kernel refactor