java调用onnx模型_ONNX预训练模型加载

tvm官网中，对从ONNX预训练模型中加载模型的教程说明

教程来自于：https://docs.tvm.ai/tutorials/frontend/from_onnx.html#sphx-glr-tutorials-frontend-from-onnx-py

首先我对教程进行了一些修改，很多东西没有必要，比如不是每次都需要从网上下载图片和模型，super_resolution.onnx和cat.png都预先下载到了文件同目录下，

同时，最新版本的tvm中不支持Python2.7，我没有编译llvm，所以我把我的设置都改到了cuda上，在24行和32行有体现,注意最新版本

1 importonnx2 importnumpy as np3 importtvm4 importtvm.relay as relay5 #from tvm.contrib.download import download_testdata

6

7 #model_url = ''.join(['https://gist.github.com/zhreshold/',

8 #'bcda4716699ac97ea44f791c24310193/raw/',

9 #'93672b029103648953c4e5ad3ac3aadf346a4cdc/',

10 #'super_resolution_0.2.onnx'])

11 #model_path = download_testdata(model_url, 'super_resolution.onnx', module='onnx')

12 #now you have super_resolution.onnx on disk

13 onnx_model = onnx.load('super_resolution.onnx')14

15 from PIL importImage16 #img_url = 'https://github.com/dmlc/mxnet.js/blob/master/data/cat.png?raw=true'

17 #img_path = download_testdata(img_url, 'cat.png', module='data')

18 img_path = 'cat.png'

19 img = Image.open(img_path).resize((224, 224))20 img_ycbcr = img.convert("YCbCr") #convert to YCbCr

21 img_y, img_cb, img_cr =img_ycbcr.split()22 x =np.array(img_y)[np.newaxis, np.newaxis, :, :]23

24 target = 'cuda'

25

26 input_name = '1'

27 shape_dict ={input_name: x.shape}28 sym, params =relay.frontend.from_onnx(onnx_model, shape_dict)29 print(sym)30

31 with relay.build_config(opt_level=1):32 intrp = relay.build_module.create_executor('graph', sym, tvm.gpu(0), target)33

34 dtype = 'float32'

35 tvm_output = intrp.evaluate(sym)(tvm.nd.array(x.astype(dtype)), **params).asnumpy()

第28行有一个从模型加载的函数from_onnx

官方的解释：tvm.relay.frontend.from_onnx(model, shape=None, dtype='float32')

Convert a ONNX model into an equivalent Relay Function.

ONNX graphs are represented as Python Protobuf objects. The companion parameters will be handled automatically. However, the input names from onnx graph is vague, mixing inputs and network weights/bias such as “1”, “2”… For convenience, we rename the real input names to “input_0”, “input_1”… And renaming parameters to “param_0”, “param_1”…

Parameters:

model (protobuf object) – ONNX ModelProto after ONNX v1.1.0

shape (dict of str to tuple,optional) – The input shape to the graph

dtype (strordict of str to str) – The input types to the graph

Returns:

sym (tvm.relay.expr.Function) – Compatible relay function

params (dict of str to tvm.NDArray) – The parameter dict to be used by relay

看返回值，sym是relay Function，在后边加一个print(sym)输出，可以看到图这一级的IR

fn (%v1: Tensor[(1, 1, 224, 224), float32], %v2: Tensor[(64, 1, 5, 5), float32], %v3: Tensor[(64,), float32], %v4: Tensor[(64, 64, 3, 3), float32], %v5: Tensor[(64,), float32], %v6: Tensor[(32, 64, 3, 3), float32], %v7: Tensor[(32,), float32], %v8: Tensor[(9, 32, 3, 3), float32], %v9: Tensor[(9,), float32]) {

%0 = nn.conv2d(%v1, %v2, padding=[2, 2], kernel_size=[5, 5])

%1 = expand_dims(%v3, axis=1, num_newaxis=2)

%2 = add(%0, %1)

%3 = nn.relu(%2)

%4 = nn.conv2d(%3, %v4, padding=[1, 1], kernel_size=[3, 3])

%5 = expand_dims(%v5, axis=1, num_newaxis=2)

%6 = add(%4, %5)

%7 = nn.relu(%6)

%8 = nn.conv2d(%7, %v6, padding=[1, 1], kernel_size=[3, 3])

%9 = expand_dims(%v7, axis=1, num_newaxis=2)

%10 = add(%8, %9)

%11 = nn.relu(%10)

%12 = nn.conv2d(%11, %v8, padding=[1, 1], kernel_size=[3, 3])

%13 = expand_dims(%v9, axis=1, num_newaxis=2)

%14 = add(%12, %13)

%15 = reshape(%14, newshape=[1, 1, 3, 3, 224, 224])

%16 = transpose(%15, axes=[0, 1, 4, 2, 5, 3])

reshape(%16, newshape=[1, 1, 672, 672])

}