Skip to content

11.22 #101

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 8 commits into
base: wlt
Choose a base branch
from
Open

11.22 #101

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
f64780c
11.22
remloveh Nov 22, 2022
edae7e1
12.6
remloveh Dec 6, 2022
bc0872c
12.6
remloveh Dec 6, 2022
f50ded8
conv2d
remloveh Dec 22, 2022
d6af4d3
op
remloveh Dec 26, 2022
1caf0ab
add DLTensor* and int32_t*
remloveh Jan 17, 2023
1bdf37b
update int32_t return value
remloveh Jan 17, 2023
cb7ae73
update op take attributes
remloveh Jan 20, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions docs/deep_dive/arch.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,9 +18,13 @@ This document is intended for developers who want to understand the architecture
process logic for specific input data with strict mathematical
description, guiding the source code.

- The :ref:`operator-and-model-test` introduces the relevant content of
operator and model testing.

.. toctree::
:maxdepth: 2

system
infer_flow
math_formalization
operator_and_model_test
253 changes: 253 additions & 0 deletions docs/deep_dive/operator_and_model_test.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,253 @@

.. _operator-and-model-test:

***********************
Operator and Model Test
***********************

.. contents::

Test Method
===========

Operator
--------

First generate test data with test_ops.py, then verify the
generated data with test_op.cc, execute the following code
to test(test_op.cc is compiled with Makefile):

.. code-block::

python tests/test_ops.py
./test_op 0/1/2 (0 is cpu, 1 is gpu, and 2 is formal)

Model
-----

First dump cvm model with main.py and corresponding yaml file, then
verify the generated model with test_model.cc, execute the following
code to test(test_model.cc is compiled with Makefile):

.. code-block::

python main.py /tests/mrt/model_zoo/yaml_file
./test_model

Test Results
============

Operator
--------

The following are the test results of the operator on different devices:

+-------------------+-------------------------+
| operator | results |
+===================+=========================+
| broadcast_sub | cpu & formal & gpu pass |
+-------------------+-------------------------+
| broadcast_add | cpu & formal & gpu pass |
+-------------------+-------------------------+
| broadcast_mul | cpu & formal & gpu pass |
+-------------------+-------------------------+
| broadcast_max | cpu & formal & gpu pass |
+-------------------+-------------------------+
| broadcast_div | cpu & formal & gpu pass |
+-------------------+-------------------------+
| broadcast_greater | cpu & formal & gpu pass |
+-------------------+-------------------------+
| max_pool2d | cpu & formal & gpu pass |
+-------------------+-------------------------+
| dense | cpu & formal & gpu pass |
+-------------------+-------------------------+
| sum | cpu & formal & gpu pass |
+-------------------+-------------------------+
| max | cpu & formal & gpu pass |
+-------------------+-------------------------+
| slice_like | cpu & formal & gpu pass |
+-------------------+-------------------------+
| tile | cpu & formal & gpu pass |
+-------------------+-------------------------+
| repeat | cpu & formal & gpu pass |
+-------------------+-------------------------+
| concatenate | cpu & formal & gpu pass |
+-------------------+-------------------------+
| transpose | cpu & formal & gpu pass |
+-------------------+-------------------------+
| take | cpu & formal & gpu pass |
+-------------------+-------------------------+
| get_valid_counts | cpu & formal & gpu pass |
+-------------------+-------------------------+
| strided_slice | cpu & formal & gpu pass |
+-------------------+-------------------------+
| conv2d | cpu & formal & gpu pass |
+-------------------|-------------------------|
| upsampling | cpu & formal & gpu pass |
+-------------------|-------------------------|


Model
-----

The following are the test results of the model on different devices:

+---------------------------+-------------------------+
| model | results |
+===========================+=========================+
| alexnet | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_resnet110_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_resnet110_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_resnet20_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_resnet20_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_resnet56_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_resnet56_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_resnext29_16x64d | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_wideresnet16_10 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_wideresnet28_10 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| cifar_wideresnet40_8 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| densenet121 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| densenet161 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| densenet169 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| densenet201 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| inceptionv1_kinetics400 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenet0.25 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenet0.5 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenet0.75 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenet1.0 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenetv2_0.25 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenetv2_0.5 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenetv2_0.75 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenetv2_1.0 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| mobilenetv3_large | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1b | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1b_kinetics400 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1c | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1d_0.73 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1d_0.76 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1d | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v1s | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet101_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet152_v1b | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet152_v1b_kinetics400 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet152_v1c | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet152_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet152_v1d | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet152_v1s | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet152_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet18_v1b_0.89 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet18_v1b | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet18_v1b_kinetics400 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet18_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet18_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet34_v1b | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet34_v1b_kinetics400 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet34_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet34_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1b | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1b_hmdb51 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1b_kinetics400 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1b_sthsthv2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1c | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1d_0.11 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1d_0.37 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1d_0.48 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1d_0.86 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1d | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v1s | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| resnet50_v2 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| squeezenet1.0 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| squeezenet1.1 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| ssd_512_mobilenet1.0_voc | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| ssd_512_resnet50_v1_voc | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg11_bn | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg11 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg13_bn | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg13 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg16_bn | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg16 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg16_ucf101 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg19_bn | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| vgg19 | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| yolo3_darknet53_voc | cpu & formal & gpu pass |
+---------------------------+-------------------------+
| yolo3_mobilenet1.0_voc | cpu & formal & gpu pass |
+---------------------------+-------------------------+
16 changes: 5 additions & 11 deletions tests/Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,17 +1,11 @@

all: test_model_cpu test_model_gpu test_model_formal test_op
all: test_model test_op

test_model_cpu: test_model.cc
g++ -o test_model_cpu test_model.cc -I../include -L../build/cpu/ -lcvm_runtime_cpu --std=c++11 -pthread -fopenmp -I../ -ldl -g -DUSE_GPU=0 -DCVM_PROFILING -fsigned-char
test_model: test_model.cc
g++ -o test_model test_model.cc -I../include -L../build/ -lcvm_runtime --std=c++11 -pthread -fopenmp -I../ -ldl -g -fsigned-char

test_model_formal: test_model.cc
g++ -o test_model_formal test_model.cc -I../include -L../build/formal/ -lcvm_runtime_formal --std=c++11 -pthread -fopenmp -I../ -ldl -g -DUSE_GPU=0 -fsigned-char

test_model_gpu: test_model.cc
g++ -o test_model_gpu test_model.cc -I../include -L../build/cpu/ -L../build/gpu/ -lcvm_runtime_cuda -lcudart -lcuda --std=c++11 -pthread -fopenmp -I./ -ldl -g -DUSE_GPU=1 -DCVM_PROFILING -fsigned-char

test_op:
g++ -o test_op test_op.cc -I../include -L../build/gpu -lcvm_runtime_cuda -lcuda -lcudart --std=c++11 -pthread -fopenmp -I../ -ldl -DCVM_PROFILING -fsigned-char -DUSE_GPU
test_op : test_op.cc
g++ -o test_op test_op.cc -I../include -L../build/ -lcvm_runtime --std=c++11 -pthread -fopenmp -I../ -ldl -g -fsigned-char

clean:
rm -f test_model_cpu test_model_formal test_model_gpu test_op
Loading