前言:记录自己安装TVM的流程,以及一个简单的利用TVM编译模型并执行的示例。
1,官网下载TVM源码
git clone --recursive https://github.com/apache/tvm
git submodule init
git submodule update
顺便完成准备工作,比如升级cmake版本需要3.18及以上版本。还有如下库:
sudo apt-get update
sudo apt-get install -y python3 python3-dev python3-setuptools gcc libtinfo-dev zlib1g-dev build-essential cmake libedit-dev libxml2-dev
2,安装clang,llvm,ninja
llvm安装依赖clang和ninja,所以直接安装llvm即可顺便完成全部的安装。
llvm ,clang安装参考:Linux系统无痛编译安装LLVM简明指南_linux安装llvm11-CSDN博客
步骤如下:
git clone git@github.com:llvm/llvm-project.git
cd llvm-project
mkdir build
cd build
sudo cmake ../llvm -DLLVM_TARGETS_TO_BUILD=X86 -DCMAKE_BUILD_TYPE=Debug
sudo make -j8
sudo make install
检查版本:
clang --version
llvm-as --version
3,安装NNPACK
NNPACK是为了优化加速神经网络的框架,可以提高在CPU上的计算效率
git clone --recursive https://github.com/Maratyszcza/NNPACK.git
cd NNPACK
# Add PIC option in CFLAG and CXXFLAG to build NNPACK shared library
sed -i "s|gnu99|gnu99 -fPIC|g" CMakeLists.txt
sed -i "s|gnu++11|gnu++11 -fPIC|g" CMakeLists.txt
mkdir build
cd build
# Generate ninja build rule and add shared library in configuration
cmake -G Ninja -D BUILD_SHARED_LIBS=ON ..
ninja
sudo ninja install
# Add NNPACK lib folder in your ldconfig
sudo sh -c "echo '/usr/local/lib'>> /etc/ld.so.conf.d/nnpack.conf"
sudo ldconfig
4,编译TVM
如下步骤,在tvm建立build文件夹,把config.cmake复制到build中
cd tvm
mkdir build
cp cmake/config.cmake build
build里的config.cmake是编译配置文件,可以按需打开关闭一些开关。下面是我修改的一些配置(TENSORRT和CUDNN我以为之前已经配置好了,结果编译报了这两个的错误,如果只是想跑流程,可以不打开这两个的开关,这样就能正常编译结束了)
set(USE_RELAY_DEBUG ON)
set(USE_CUDA ON)
set(USE_NNPACK ON)
set(USE_LLVM ON)
set(USE_TENSORRT_CODEGEN ON)
set(USE_TENSORRT_RUNTIME ON)
set(USE_CUDNN ON)
编译代码:
cd build
cmake ..
make -j12
5,配置python环境
从build文件夹出来进入到tvm/python文件夹下,执行如下命令,即可配置python中的tvm库了。
cd ../python
python setup.py install
python中使用tvm测试,导入tvm不出错即配置tvm安装成功
import tvm
print(tvm.__version__)
6,一个简单示例
该测试来自TVM官方文档的示例,包括编译一个测试执行一个分类网络和编译器自动调优测试。仅先直观的看到TVM如何作为一个工具对模型编译并部署的流程。
1) 下载onnx模型
wget https://github.com/onnx/models/raw/b9a54e89508f101a1611cd64f4ef56b9cb62c7cf/vision/classification/resnet/model/resnet50-v2-7.onnx
2) 编译onnx模型
python -m tvm.driver.tvmc compile --target "llvm" --input-shapes "data:[1,3,224,224]" --output resnet50-v2-7-tvm.tar resnet50-v2-7.onnx
如果报这样的警告:
就在git上下载一份tophub,把整个文件夹tophub复制到 ~/.tvm/路径下
git clone git@github.com:tlc-pack/tophub.git
sudo cp -r tophub ~/.tvm/
解压生成的tvm编译模型,得到3个文件:
-
mod.so
作为一个C++库的编译模型, 能被 TVM runtime加载 -
mod.json TVM Relay计算图的文本表示
-
mod.params
onnx模型的预训练权重参数
mkdir model
tar -xvf resnet50-v2-7-tvm.tar -C model
ls model
3) 输入数据前处理
python preprocess.py
图像处理代码文件:preprocess.py
#!python ./preprocess.py
from tvm.contrib.download import download_testdata
from PIL import Image
import numpy as np
img_url = "https://s3.amazonaws.com/model-server/inputs/kitten.jpg"
img_path = download_testdata(img_url, "imagenet_cat.png", module="data")
# Resize it to 224x224
resized_image = Image.open(img_path).resize((224, 224))
img_data = np.asarray(resized_image).astype("float32")
# ONNX expects NCHW input, so convert the array
img_data = np.transpose(img_data, (2, 0, 1))
# Normalize according to ImageNet
imagenet_mean = np.array([0.485, 0.456, 0.406])
imagenet_stddev = np.array([0.229, 0.224, 0.225])
norm_img_data = np.zeros(img_data.shape).astype("float32")
for i in range(img_data.shape[0]):
norm_img_data[i, :, :] = (img_data[i, :, :] / 255 - imagenet_mean[i]) / imagenet_stddev[i]
# Add batch dimension
img_data = np.expand_dims(norm_img_data, axis=0)
# Save to .npz (outputs imagenet_cat.npz)
np.savez("imagenet_cat", data=img_data)
4) 运行编译模型
python -m tvm.driver.tvmc run --inputs imagenet_cat.npz --output predictions.npz resnet50-v2-7-tvm.tar
5) 输出后处理
python postprocess.py
执行之后得到分类结果的输出:
class='n02123045 tabby, tabby cat' with probability=0.621104
class='n02123159 tiger cat' with probability=0.356378
class='n02124075 Egyptian cat' with probability=0.019712
class='n02129604 tiger, Panthera tigris' with probability=0.001215
class='n04040759 radiator' with probability=0.000262
后处理代码:postprocess.py
#!python ./postprocess.py
import os.path
import numpy as np
from scipy.special import softmax
from tvm.contrib.download import download_testdata
# Download a list of labels
labels_url = "https://s3.amazonaws.com/onnx-model-zoo/synset.txt"
labels_path = download_testdata(labels_url, "synset.txt", module="data")
with open(labels_path, "r") as f:
labels = [l.rstrip() for l in f]
output_file = "predictions.npz"
# Open the output and read the output tensor
if os.path.exists(output_file):
with np.load(output_file) as data:
scores = softmax(data["output_0"])
scores = np.squeeze(scores)
ranks = np.argsort(scores)[::-1]
for rank in ranks[0:5]:
print("class='%s' with probability=%f" % (labels[rank], scores[rank]))
6) 编译器自动调优
调优的算法使用的是xgboost,所以需要python安装一下这个库。
pip install xgboost
python -m tvm.driver.tvmc tune --target "llvm" --output resnet50-v2-7-autotuner_records.json resnet50-v2-7.onnx
7) 重新编译并执行调优后的模型
python -m tvm.driver.tvmc compile --target "llvm" --tuning-records resnet50-v2-7-autotuner_records.json --output resnet50-v2-7-tvm_autotuned.tar resnet50-v2-7.onnx
python -m tvm.driver.tvmc run --inputs imagenet_cat.npz --output predictions.npz resnet50-v2-7-tvm_autotuned.tar
python postprocess.py
预测结果:
class='n02123045 tabby, tabby cat' with probability=0.610552
class='n02123159 tiger cat' with probability=0.367180
class='n02124075 Egyptian cat' with probability=0.019365
class='n02129604 tiger, Panthera tigris' with probability=0.001273
class='n04040759 radiator' with probability=0.000261
8) 比较编译前后执行模型的速度
python -m tvm.driver.tvmc run --inputs imagenet_cat.npz --output predictions.npz --print-time --repeat 100 resnet50-v2-7-tvm_autotuned.tar
python -m tvm.driver.tvmc run --inputs imagenet_cat.npz --output predictions.npz --print-time --repeat 100 resnet50-v2-7-tvm.tar
执行时间如下,上面是自动调优过的的,可以明显看出推理时间上的优化效果。
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
84.6208 74.9435 143.9276 72.8249 19.0734
mean (ms) median (ms) max (ms) min (ms) std (ms)
131.1953 130.7819 140.6614 106.0725 3.5606
比较了一下两个编译后模型的Relay计算图json文件的区别,就看到了算子数据layout的区别,更多细节还是要看源码吧
参考:TVM Ubuntu20安装_ubuntu20.04配置tvm_shelgi的博客-CSDN博客