[NNAdapter][EEasyTechNPU] Init support for EEasyTechNPU #8960
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
yingshengBD
requested review from
hp03,
chenjiaoAngel and
hong19860320
as code owners
|
Thanks for your contribution! |
yingshengBD
changed the title
hong19860320
reviewed
May 4, 2022
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/CMakeLists.txt
Show resolved
Hide resolved
hong19860320
reviewed
May 4, 2022
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/CMakeLists.txt
Show resolved
Hide resolved
| input_tensor = converter->ConvertOperand(input_operand); | ||
| } | ||
| auto output_tensor = converter->ConvertOperand(output_operand); | ||
| auto new_scale_operand = converter->GetOperand(scale_operand); |
There was a problem hiding this comment.
可以去掉GetOperand和ClearOperands接口，直接使用临时变量存储转换后的bias 、scale数据，然后再调用AddXXXConstantTensor就行了，如果eeasy ddk只支持引用buffer，可以在Converter维护一个std::vector<std::vector<int8_t>> buffers_; 在完成模型生成后，随着Converter的销毁一起销毁、
There was a problem hiding this comment.
在新commit中修复，使用AddVariableTensor和AddConstantTensor将BN转换成mul和add
hong19860320
reviewed
May 4, 2022
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/converter/converter.cc
Show resolved
Hide resolved
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/converter/converter.cc
Show resolved
Hide resolved
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/converter/converter.cc
Outdated
Show resolved
Hide resolved
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/converter/converter.cc
Outdated
Show resolved
Hide resolved
| // Convert a NNAdapter model to eznn graph and tensors | ||
| int Apply(core::Model* model); | ||
| // Mapping a eznn tensor to a NNAdapter operand | ||
| std::string GetTensorName(core::Operand* operand); |
There was a problem hiding this comment.
已删除，在新commit中修复
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/converter/resize_linear.cc
Show resolved
Hide resolved
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/converter/resize_nearest.cc
Show resolved
Hide resolved
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/converter/split.cc
Outdated
Show resolved
Hide resolved
| #add_library(eznpu_ddk SHARED IMPORTED GLOBAL) | ||
| #set_property(TARGET eznpu_ddk PROPERTY IMPORTED_LOCATION libeznpu_ddk.so) | ||
|
|
||
| set(${DEVICE_NAME}_deps eznpu_ddk) |
There was a problem hiding this comment.
根据其它硬件的最新版dependencies.cmake改下
lite/backends/nnadapter/nnadapter/src/driver/eeasytech_npu/utility.h
Outdated
Show resolved
Hide resolved
Comment on lines
18
to
30
| add_definitions(-DNNADAPTER_DEVICE_SYMBOL=${NNADAPTER_DEVICE_SYMBOL_PREFIX}${DEVICE_NAME}) | ||
|
|
||
| include(dependencies.cmake) | ||
|
|
||
| include_directories(.) | ||
| aux_source_directory(converter CONVERTERS) | ||
| aux_source_directory(optimizer OPTIMIZERS) | ||
| set(SRCS utility.cc ${OPTIMIZERS} ${CONVERTERS} engine.cc driver.cc) | ||
| set(DEPS ${NNADAPTER_OPERATIONS} ${NNADAPTER_UTILITIES} nnadapter_optimizer_convert_quantization_symm_to_asymm nnadapter_optimizer_fuse_matmul_add_into_fully_connected ${${DEVICE_NAME}_deps}) | ||
|
|
||
| add_library(${DEVICE_NAME} SHARED ${SRCS}) | ||
| target_link_libraries(${DEVICE_NAME} "-Wl,--start-group" ${DEPS} "-Wl,--end-group") | ||
| set(NNADAPTER_DEVICES ${NNADAPTER_DEVICES} ${DEVICE_NAME} CACHE INTERNAL "") |
There was a problem hiding this comment.
还是没改回来， 是全部要改，可以参考下
|
|
||
| int ConvertElementwise(Converter* converter, core::Operation* operation) { | ||
| ELEMENTWISE_OPERATION_EXTRACT_INPUTS_OUTPUTS | ||
| // Convert to eeasynpu tensors and operators |
| int ConvertResizeNearest(Converter *converter, core::Operation *operation) { | ||
| RESIZE_NEAREST_OPERATION_EXTRACT_INPUTS_OUTPUTS | ||
| NNADAPTER_CHECK_EQ(IsOperandWithDynamicShape(input_operand), false); | ||
| // Convert to eeasynpu tensors and node |
| int ConvertResizeLinear(Converter *converter, core::Operation *operation) { | ||
| RESIZE_LINEAR_OPERATION_EXTRACT_INPUTS_OUTPUTS | ||
| NNADAPTER_CHECK_EQ(IsOperandWithDynamicShape(input_operand), false); | ||
| // Convert to eeasynpu tensors and node |
There was a problem hiding this comment.
空一行？所有converter 的 // Convert to eeasynpu tensors and node 前面都空一行吧
|
|
||
| int ConvertResizeLinear(Converter *converter, core::Operation *operation) { | ||
| RESIZE_LINEAR_OPERATION_EXTRACT_INPUTS_OUTPUTS | ||
| NNADAPTER_CHECK_EQ(IsOperandWithDynamicShape(input_operand), false); |
There was a problem hiding this comment.
直接 NNADAPTER_CHECK(!IsOperandWithDynamicShape(input_operand)); 就行了
|
|
||
| NNADAPTER_EXPORT nnadapter::driver::Device NNADAPTER_AS_SYM2(DEVICE_NAME) = { | ||
| .name = NNADAPTER_AS_STR2(DEVICE_NAME), | ||
| .vendor = "Eeasytech", |
There was a problem hiding this comment.
vendor 名称和亿智确认下吧，使用 EEasyTech 、eeasytech 还是 EEASY-TECH
There was a problem hiding this comment.
嗯，已经和亿智相关同学确认，为全大写EEASYTECH，在新commit中修复
test=develop
test=develop
test=develop
test=develop
test=develop
test=develop
hong19860320
changed the title
hong19860320
changed the title
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.
init support for eeasytechNPU
SoC: eeasytech 806/810，506/510，206/210
cotain OPs：add, mul, sub, avg_pool, max_pool, bn, conv2d, flatten, reshape, fc, hard-swish, sigmoid, tanh, split, concat, interp_bilinear, interp_nearest, relu, relu6, transpose