From 17b622c8825e6a85f7d94275dff971d53ec2baa8 Mon Sep 17 00:00:00 2001 From: Guanzhong Chen Date: Fri, 15 Dec 2023 11:12:34 +0800 Subject: [PATCH 1/2] 1 --- .../built-in/cv/gan/stylegan2-ada/README.md | 171 ++++++++++++++++++ .../built-in/cv/gan/stylegan2-ada/export.py | 92 ++++++++++ .../built-in/cv/gan/stylegan2-ada/perf.py | 92 ++++++++++ .../built-in/cv/gan/stylegan2-ada/perf_gpu.py | 125 +++++++++++++ .../cv/gan/stylegan2-ada/requirements.txt | 2 + .../stylegan2-ada-pytorch_postprocess.py | 98 ++++++++++ .../stylegan2-ada-pytorch_preprocess.py | 82 +++++++++ 7 files changed, 662 insertions(+) create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/requirements.txt create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_postprocess.py create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_preprocess.py diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md new file mode 100644 index 0000000000..c7f6f4652f --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md @@ -0,0 +1,171 @@ +# StyleGAN2-ADA模型-推理指导 + + +- [概述](#ZH-CN_TOPIC_0000001172161501) + + - [输入输出数据](#section540883920406) + + + +- [推理环境准备](#ZH-CN_TOPIC_0000001126281702) + +- [快速上手](#ZH-CN_TOPIC_0000001126281700) + + - [获取源码](#section4622531142816) + - [准备数据集](#section183221994411) + - [模型推理](#section741711594517) + +- [模型推理性能&精度](#ZH-CN_TOPIC_0000001172201573) + + ****** + +# 概述 + +StyleGAN2-ADA是具有自适应鉴别器增强（ADA）的StyleGAN2，用有限的数据训练生成对抗网络，[论文链接](https://arxiv.org/abs/2006.06676)。 + +- 参考实现： + + ``` + url=https://github.com/NVlabs/stylegan2-ada-pytorch + commit_id=765125e7f01a4c265e25d0a619c37a89ac3f5107 + code_path= + model_name=StyleGAN2-ADA + ``` + +## 输入输出数据 + +- 输入数据 + + | 输入数据 | 数据类型 | 大小 | 数据排布格式 | + | -------- | -------- | ------------------------- | ------------ | + | input | RGB_FP32 | batchsize x 512 | ND | + + +- 输出数据 + + | 输出数据 | 数据类型 | 大小 | 数据排布格式 | + | -------- | -------- | -------- | ------------ | + | output1 | FLOAT32 | batchsize x 3 x 512 x 512| NCHW | + + +# 推理环境准备 + +- 该模型需要以下插件与驱动 + + **表 1** 版本配套表 + + | 配套 | 版本 | 环境准备指导 | + |---------| ------- | ------------------------------------------------------------ | + | 固件与驱动 | 23.0.rc1 | [Pytorch框架推理环境准备](https://www.hiascend.com/document/detail/zh/ModelZoo/pytorchframework/pies) | + | CANN | 7.0.RC1.alpha003 | - | + | Python | 3.9.11 | - | + | PyTorch | 2.0.1 | - | + | Torch_AIE | 6.3.rc2 | - | + +# 快速上手 + +## 获取源码 + +1. 获取源码。 + + ``` + git clone https://github.com/NVlabs/stylegan2-ada-pytorch.git + ``` + +2. 安装依赖。 + + ``` + pip3 install -r requirements.txt + ``` + +## 准备数据集 + +1. 获取原始数据集。（解压命令参考tar –xvf \*.tar与 unzip \*.zip） + + StyleGAN2-ADA网络使用随机生成隐变量作为输入来生成输出，生成方式见下一步。 + + +2. 数据预处理，将原始数据集转换为模型输入的数据。 + + 执行stylegan2-ada-pytorch_preprocess.py脚本，完成预处理。 + + ``` + python3 stylegan2-ada-pytorch_preprocess.py --num_input=200 --save_path=./pre_data + ``` + 生成`num_input个`随机输入，并保存为bin文件，保存目录为`./pre_data`。 + + +## 模型推理 + +1. 开始推理验证。 + + 1. 获取权重文件。 + + 权重文件为：[G_ema_bs8_8p_kimg1000.pkl](https://ascend-repo-modelzoo.obs.cn-east-2.myhuaweicloud.com/model/1_PyTorch_PTH/StyleGAN/PTH/G_ema_bs8_8p_kimg1000.pkl) + 将获取的权重文件放在当前工作路径下。 + + 2. 导出ts模型 + + 执行如下命令导出ts模型文件： + ``` + python3 export.py --stylegan_path ./stylegan2-ada-pytorch-main --pkl_file ./G_ema_bs8_8p_kimg1000.pkl + ``` + - 参数说明： + + - stylegan_path：拉下github仓的目录 + - pkl_file：pkl文件目录 + + + 3. 执行推理，生成伪装图片 + + ``` + python3 stylegan2-ada-pytorch_postprocess.py --bin_path ./pre_data --image_path ./results --ts_model_path ./stylegan2.ts + ``` + + - 参数说明： + + - bin_path：预处理得到的bin文件的目录 + - image_path：推理后得到图片的目录 + - ts_model_path：ts模型文件路径 + + + 3. 精度验证（肉眼验证）。 + + a.调用`perf_gpu.py`脚本使用pkl权重文件成生图像，用于进行精度对比。 + + ``` + python3 perf_gpu.py --stylegan_path ./stylegan2-ada-pytorch --pkl_file ./G_ema_bs8_8p_kimg1000.pkl --input_path ./pre_data --image_path ./results + ``` + + - 参数说明： + + - stylegan_path：拉下github仓的目录 + - pkl_file：pkl文件目录 + - input_path：预处理得到的bin文件的目录 + - image_path：推理后得到图片的目录 + + b.精度比对方法：将ts模型的推理结果转化的图像与pkl权重文件成生图像进行对比（均在image path路径下，原pkl推理结果在pkl_img子目录下），两幅图像在视觉效果上一致即可。 + + + 4. 性能验证。 + + 使用`perf.py`脚本进行性能验证 + ``` + python3 perf.py --mode ts --ts_path ./stylegan2.ts --batch_size 1 + ``` + + - 参数说明： + + - mode：使用ts模型进行推理 + - ts_path：ts模型文件所在路径 + - batch_size：用于验证性能的batch size路径 + + +# 模型推理性能&精度 + +调用ACL接口推理计算，性能参考下列数据。 + +| 芯片型号 | Batch Size | 数据集 | 精度 | 性能 | +| --------- | ---------------- | ---------- | ---------- | --------------- | +| Ascend310P3 | 1 | 随机生成数据 | 图片视觉评价 | 23.50 fps | +| Ascend310P3 | 16 | 随机生成数据 | 图片视觉评价 | 24.59 fps | \ No newline at end of file diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py new file mode 100644 index 0000000000..e8bd32cbe5 --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py @@ -0,0 +1,92 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the License); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# BSD 3-Clause License +# +# Copyright (c) 2017 xxxx +# All rights reserved. +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# +# * Neither the name of the copyright holder nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# ============================================================================ +import sys +import pickle +import argparse +import functools + +import torch + + +def parse_args(): + args = argparse.ArgumentParser(description="A program that operates in 'om' or 'ts' mode.") + args.add_argument('--stylegan_path', type=str, default='./stylegan2-ada-pytorch-main', + help='Path to the stylegan2-ada-pytorch Github directory') + args.add_argument('--pkl_file',help='stylegan2-ada pth file path', type=str, + default='./G_ema_bs8_8p_kimg1000.pkl' + ) + return args.parse_args() + + +def pkl2onnx(args): + sys.path.append(args.stylegan_path) + + # Set up options + bs = 16 + pkl_file = args.pkl_file + + # Load pkl model + with open(pkl_file, 'rb') as f: + G = pickle.load(f)['G_ema'] + G.forward = functools.partial(G.forward, force_fp32=True) + + print("type of G is:", type(G)) + + # Prepare input + print(G.z_dim) + z = torch.randn([bs, G.z_dim]) + c = torch.empty([bs, 0], device='cpu') + dummy_input = (z, c) + G.eval() + + ts_model = torch.jit.trace(G, dummy_input) + ts_model.save("./stylegan2-bs16.ts") + + +if __name__ == "__main__": + args = parse_args() + pkl2onnx(args) diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py new file mode 100644 index 0000000000..b15cfaee08 --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py @@ -0,0 +1,92 @@ +import argparse +import time +from tqdm import tqdm + +import torch +import numpy as np + +import torch_aie + + +INPUT_WIDTH = 512 + +def parse_args(): + args = argparse.ArgumentParser(description="A program that operates in 'om' or 'ts' mode.") + args.add_argument("--mode", choices=["om", "ts"], required=True, help="Specify the mode ('om' or 'ts').") + args.add_argument('--ts_path',help='stylegan2 ts file path', type=str, + default='./stylegan2-bs16.ts' + ) + args.add_argument("--batch_size", type=int, default=16, help="batch size.") + return args.parse_args() + + +if __name__ == '__main__': + infer_times = 100 + om_cost = 0 + pt_cost = 0 + opts = parse_args() + batch_size = opts.batch_size + + if opts.mode == "om": + om_model = InferSession(0, OM_PATH) + for _ in tqdm(range(0, infer_times)): + dummy_input = np.random.randn(1, 3, INPUT_WIDTH, INPUT_HEIGHT).astype(np.uint8) + start = time.time() + output = om_model.infer([dummy_input], 'static', custom_sizes=90000000) # revise static + # output = om_model.infer([dummy_input], 'dymshape', custom_sizes=4000) # revise dynm fp32为4个字节，输出为1x1000 + cost = time.time() - start + om_cost += cost + + if opts.mode == "ts": + ts_model = torch.jit.load(opts.ts_path) + + input_info = [torch_aie.Input((batch_size, INPUT_WIDTH)), torch_aie.Input((batch_size, 0))] + + torch_aie.set_device(0) + print("start compile") + torchaie_model = torch_aie.compile( + ts_model, + inputs=input_info, + precision_policy=torch_aie.PrecisionPolicy.PREF_FP32, + soc_version='Ascend310P3', + ) + print("end compile") + torchaie_model.eval() + + input = np.zeros((batch_size, 512)) + input = torch.Tensor(input) + c = torch.empty([batch_size, 0]) + input_tensor = input.to('npu') + input_tensor2 = c.to('npu') + print("ready to infer") + _ = torchaie_model(input_tensor, input_tensor2) + print("finish infer once") + loops = 100 + warm_ctr = 10 + + default_stream = torch_aie.npu.default_stream() + time_cost = 0 + + while warm_ctr: + _ = torchaie_model(input_tensor, input_tensor2) + default_stream.synchronize() + warm_ctr -= 1 + + for i in range(loops): + t0 = time.time() + _ = torchaie_model(input_tensor, input_tensor2) + default_stream.synchronize() + t1 = time.time() + time_cost += (t1 - t0) + print(i) + + print(f"fps: {loops} * {batch_size} / {time_cost : .3f} samples/s") + print("torch_aie fps: ", loops * batch_size / time_cost) + + from datetime import datetime + current_time = datetime.now() + formatted_time = current_time.strftime("%Y-%m-%d %H:%M:%S") + print("Current Time:", formatted_time) + + + diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py new file mode 100644 index 0000000000..0fbcaf6eff --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py @@ -0,0 +1,125 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the License); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# +# BSD 3-Clause License +# +# Copyright (c) 2017 xxxx +# All rights reserved. +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# +# * Neither the name of the copyright holder nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# ============================================================================ + +import os +import sys +import time +import pickle +import argparse +import functools +from tqdm import tqdm + +import numpy as np +import torch +import PIL.Image + + +def save_image_grid(img, fname, drange, grid_size): + lo, hi = drange + img = np.asarray(img, dtype=np.float32) + img = (img - lo) * (255 / (hi - lo)) + img = np.rint(img).clip(0, 255).astype(np.uint8) + + gw, gh = grid_size + _N, C, H, W = img.shape + img = img.reshape(gh, gw, C, H, W) + img = img.transpose(0, 3, 1, 4, 2) + img = img.reshape(gh * H, gw * W, C) + + assert C in [1, 3] + if C == 1: + PIL.Image.fromarray(img[:, :, 0], 'L').save(fname) + if C == 3: + PIL.Image.fromarray(img, 'RGB').save(fname) + + +def main(args): + pkl_file = args.pkl_file + bs = args.batch_size + input_path = args.input_path + image_path = args.image_path + device = "cpu" + + grid_size = (1, 1) + input_files = os.listdir(input_path) + input_files.sort() + image_path = os.path.join(image_path, 'pkl_img') + os.makedirs(image_path, exist_ok=True) + # load model + start = time.time() + with open(pkl_file, 'rb') as f: + G = pickle.load(f)['G_ema'] + + G.forward = functools.partial(G.forward, force_fp32=True) + for i in tqdm(range(len(input_files))): + input_file = input_files[i] + input_file = os.path.join(input_path, input_file) + input_file = np.fromfile(input_file, dtype=np.float32) + z = torch.tensor(input_file).reshape(-1, G.z_dim).to(device) + c = torch.empty(bs, 0).to(device) + image = G(z, c) + image = image.reshape(-1, 3, 512, 512) + image = image.cpu() + save_image_grid(image, os.path.join(image_path, f'gen_image_{i:04d}') + '.png', drange=[-1, 1], + grid_size=grid_size) + + end = time.time() + print(f'Inference average time : {((end - start) * 1000 / len(input_files)):.2f} ms') + + +if __name__ == '__main__': + parser = argparse.ArgumentParser() + parser.add_argument('--stylegan_path', type=str, default='./stylegan2-ada-pytorch', + help='Path to the stylegan2-ada-pytorch Github directory') + parser.add_argument('--pkl_file', type=str, default='./G_ema_bs8_8p_kimg1000.pkl') + parser.add_argument('--input_path', type=str, default='./pre_data') + parser.add_argument('--image_path', type=str, default='./results') + parser.add_argument('--batch_size', type=int, default=1) + args = parser.parse_args() + + sys.path.append(args.stylegan_path) + + main(args) diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/requirements.txt b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/requirements.txt new file mode 100644 index 0000000000..ed86fa1b6e --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/requirements.txt @@ -0,0 +1,2 @@ +torch==2.0.1 +pillow==10.1.0 \ No newline at end of file diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_postprocess.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_postprocess.py new file mode 100644 index 0000000000..15e1d00203 --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_postprocess.py @@ -0,0 +1,98 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the License); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# BSD 3-Clause License +# +# Copyright (c) 2017 xxxx +# All rights reserved. +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# +# * Neither the name of the copyright holder nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# ============================================================================ + +import torch +import os +import numpy as np +import argparse + +from tqdm import tqdm +from perf_gpu import save_image_grid + +import torch_aie + + +def test_om(args): + ts_model_path = args.ts_model_path + ts_model = torch.jit.load(ts_model_path) + input_info = [torch_aie.Input((1, 512)), torch_aie.Input((1, 0))] + torch_aie.set_device(0) + print("start_compile") + torchaie_model = torch_aie.compile( + ts_model, + inputs=input_info, + precision_policy=torch_aie.PrecisionPolicy.PREF_FP32, # _enums.PrecisionPolicy.FP32 + soc_version='Ascend310P3' + ) + print("end_compile") + torchaie_model.eval() + + bin_path = args.bin_path + image_path = args.image_path + bin_list = os.listdir(bin_path) + bin_list.sort() + for i in tqdm(range(len(bin_list))): + input = np.fromfile(os.path.join(bin_path, bin_list[i]), dtype=np.float32).reshape(-1,512) + input = torch.Tensor(input) + c = torch.empty([1, 0]) + images = torchaie_model(input.to('npu'), c.to('npu')) + images = images.to('cpu') + bs = images.shape[0] + grid_size = (4, 4) if bs == 16 else (1, 1) + os.makedirs(image_path, exist_ok=True) + save_image_grid(images, os.path.join(image_path, f'gen_image_{i:04d}') + ".png", drange=[-1, 1], + grid_size=grid_size) + + +if __name__ == "__main__": + parser = argparse.ArgumentParser() + parser.add_argument('--bin_path', type=str, default="./pre_data") + parser.add_argument('--image_path', type=str, default='./results') + parser.add_argument('--ts_model_path', type=str, default='./stylegan2.ts') + args = parser.parse_args() + + test_om(args) diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_preprocess.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_preprocess.py new file mode 100644 index 0000000000..cc2fd5cac9 --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_preprocess.py @@ -0,0 +1,82 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the License); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# BSD 3-Clause License +# +# Copyright (c) 2017 xxxx +# All rights reserved. +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# +# * Neither the name of the copyright holder nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# ============================================================================ + +import os +import torch +import argparse + +from tqdm import tqdm + + +def main(args): + # set up option + z_dim = 512 + c_dim = 0 + bs = 1 + num = args.num_input + save_path = args.save_path + + # create save path dir + os.makedirs(save_path, exist_ok=True) + + # generate input + for i in tqdm(range(num)): + z = torch.randn([bs, z_dim]) + c = torch.empty([bs, c_dim]) + input = torch.cat((z, c), 1).numpy() + input.tofile(os.path.join(save_path, f'input_bs{bs}_{i:04d}.bin')) + + +# ---------------------------------------------------------------------------- +if __name__ == "__main__": + parser = argparse.ArgumentParser() + parser.add_argument('--num_input', type=int, default=1) + parser.add_argument('--batch_size', type=int, default=1) + parser.add_argument('--save_path', type=str, default='./input') + args = parser.parse_args() + main(args) +# ---------------------------------------------------------------------------- -- Gitee From f01560a82e7049ace9c8c6bd51bd38e7bb11da0c Mon Sep 17 00:00:00 2001 From: Guanzhong Chen Date: Tue, 26 Dec 2023 21:55:05 +0800 Subject: [PATCH 2/2] =?UTF-8?q?=E6=96=B0=E5=A2=9E=E6=B5=8B=E8=AF=95?= =?UTF-8?q?=E7=94=A8=E4=BE=8B?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../built-in/cv/gan/stylegan2-ada/README.md | 13 +- .../built-in/cv/gan/stylegan2-ada/export.py | 23 +++- ...n2-ada-pytorch_postprocess.py => infer.py} | 0 .../built-in/cv/gan/stylegan2-ada/perf.py | 118 ++++++++--------- .../built-in/cv/gan/stylegan2-ada/perf_gpu.py | 2 +- ...da-pytorch_preprocess.py => preprocess.py} | 26 ++-- .../cv/gan/stylegan2-ada/test_export.py | 119 ++++++++++++++++++ .../cv/gan/stylegan2-ada/test_infer.py | 117 +++++++++++++++++ .../cv/gan/stylegan2-ada/test_perf.py | 91 ++++++++++++++ 9 files changed, 424 insertions(+), 85 deletions(-) rename AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/{stylegan2-ada-pytorch_postprocess.py => infer.py} (100%) rename AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/{stylegan2-ada-pytorch_preprocess.py => preprocess.py} (80%) create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_export.py create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_infer.py create mode 100644 AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_perf.py diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md index c7f6f4652f..d9e7975c86 100644 --- a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/README.md @@ -16,6 +16,7 @@ - [模型推理](#section741711594517) - [模型推理性能&精度](#ZH-CN_TOPIC_0000001172201573) +- [开发者自测 && 测试覆盖率](#section741711594518) ****** @@ -87,10 +88,10 @@ StyleGAN2-ADA是具有自适应鉴别器增强（ADA）的StyleGAN2，用有限 2. 数据预处理，将原始数据集转换为模型输入的数据。 - 执行stylegan2-ada-pytorch_preprocess.py脚本，完成预处理。 + 执行`preprocess.py`脚本，完成预处理。 ``` - python3 stylegan2-ada-pytorch_preprocess.py --num_input=200 --save_path=./pre_data + python3 preprocess.py --num_input=200 --save_path=./pre_data ``` 生成`num_input个`随机输入，并保存为bin文件，保存目录为`./pre_data`。 @@ -119,7 +120,7 @@ StyleGAN2-ADA是具有自适应鉴别器增强（ADA）的StyleGAN2，用有限 3. 执行推理，生成伪装图片 ``` - python3 stylegan2-ada-pytorch_postprocess.py --bin_path ./pre_data --image_path ./results --ts_model_path ./stylegan2.ts + python3 infer.py --bin_path ./pre_data --image_path ./results --ts_model_path ./stylegan2.ts ``` - 参数说明： @@ -168,4 +169,8 @@ StyleGAN2-ADA是具有自适应鉴别器增强（ADA）的StyleGAN2，用有限 | 芯片型号 | Batch Size | 数据集 | 精度 | 性能 | | --------- | ---------------- | ---------- | ---------- | --------------- | | Ascend310P3 | 1 | 随机生成数据 | 图片视觉评价 | 23.50 fps | -| Ascend310P3 | 16 | 随机生成数据 | 图片视觉评价 | 24.59 fps | \ No newline at end of file +| Ascend310P3 | 16 | 随机生成数据 | 图片视觉评价 | 24.59 fps | + +# 开发者自测 && 测试覆盖率 + +![](coverage.png) \ No newline at end of file diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py index e8bd32cbe5..76ec6a26d3 100644 --- a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/export.py @@ -1,4 +1,4 @@ -# Copyright 2021 Huawei Technologies Co., Ltd +# Copyright 2023 Huawei Technologies Co., Ltd # # Licensed under the Apache License, Version 2.0 (the License); # you may not use this file except in compliance with the License. @@ -44,6 +44,9 @@ # OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. # ============================================================================ + + +import os import sys import pickle import argparse @@ -53,7 +56,7 @@ import torch def parse_args(): - args = argparse.ArgumentParser(description="A program that operates in 'om' or 'ts' mode.") + args = argparse.ArgumentParser(description="A program that trace the pkl model into torchscript file") args.add_argument('--stylegan_path', type=str, default='./stylegan2-ada-pytorch-main', help='Path to the stylegan2-ada-pytorch Github directory') args.add_argument('--pkl_file',help='stylegan2-ada pth file path', type=str, @@ -62,11 +65,18 @@ def parse_args(): return args.parse_args() -def pkl2onnx(args): +def check_args(args): + if not os.path.exists(args.stylegan_path): + raise FileNotFoundError(f'The stylegan2-ada-pytorch Github directory {args.stylegan_path} not exists') + if not os.path.exists(args.pkl_file): + raise FileNotFoundError(f'The model file stylegan2-ada pth {args.pkl_file} not exists') + + +def trace_ts_model(args): sys.path.append(args.stylegan_path) # Set up options - bs = 16 + bs = 1 pkl_file = args.pkl_file # Load pkl model @@ -84,9 +94,10 @@ def pkl2onnx(args): G.eval() ts_model = torch.jit.trace(G, dummy_input) - ts_model.save("./stylegan2-bs16.ts") + ts_model.save("./stylegan2-bs.ts") if __name__ == "__main__": args = parse_args() - pkl2onnx(args) + check_args(args) + trace_ts_model(args) diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_postprocess.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/infer.py similarity index 100% rename from AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_postprocess.py rename to AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/infer.py diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py index b15cfaee08..4765be3632 100644 --- a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf.py @@ -1,87 +1,56 @@ import argparse import time -from tqdm import tqdm import torch import numpy as np import torch_aie +from torch_aie import _enums -INPUT_WIDTH = 512 +INPUT_SIZE = 512 + def parse_args(): - args = argparse.ArgumentParser(description="A program that operates in 'om' or 'ts' mode.") - args.add_argument("--mode", choices=["om", "ts"], required=True, help="Specify the mode ('om' or 'ts').") + args = argparse.ArgumentParser(description="A program that gets the performance of torchaie on stylegan.") args.add_argument('--ts_path',help='stylegan2 ts file path', type=str, - default='./stylegan2-bs16.ts' + default='./stylegan2-bs.ts' ) - args.add_argument("--batch_size", type=int, default=16, help="batch size.") + args.add_argument("--batch_size", type=int, default=1, help="batch size.") + args.add_argument('--image_size', type=int, default=512, help='Image size') return args.parse_args() -if __name__ == '__main__': - infer_times = 100 - om_cost = 0 - pt_cost = 0 - opts = parse_args() - batch_size = opts.batch_size - - if opts.mode == "om": - om_model = InferSession(0, OM_PATH) - for _ in tqdm(range(0, infer_times)): - dummy_input = np.random.randn(1, 3, INPUT_WIDTH, INPUT_HEIGHT).astype(np.uint8) - start = time.time() - output = om_model.infer([dummy_input], 'static', custom_sizes=90000000) # revise static - # output = om_model.infer([dummy_input], 'dymshape', custom_sizes=4000) # revise dynm fp32为4个字节，输出为1x1000 - cost = time.time() - start - om_cost += cost - - if opts.mode == "ts": - ts_model = torch.jit.load(opts.ts_path) - - input_info = [torch_aie.Input((batch_size, INPUT_WIDTH)), torch_aie.Input((batch_size, 0))] - - torch_aie.set_device(0) - print("start compile") - torchaie_model = torch_aie.compile( - ts_model, - inputs=input_info, - precision_policy=torch_aie.PrecisionPolicy.PREF_FP32, - soc_version='Ascend310P3', - ) - print("end compile") - torchaie_model.eval() - - input = np.zeros((batch_size, 512)) - input = torch.Tensor(input) - c = torch.empty([batch_size, 0]) - input_tensor = input.to('npu') - input_tensor2 = c.to('npu') - print("ready to infer") +def perf(torchaie_model, batch_size, image_size): + input = np.zeros((batch_size, image_size)) + input = torch.Tensor(input) + c = torch.empty([batch_size, 0]) + input_tensor = input.to('npu') + input_tensor2 = c.to('npu') + print("ready to infer") + _ = torchaie_model(input_tensor, input_tensor2) + print("finish infer once") + loops = 100 + warm_ctr = 10 + + default_stream = torch_aie.npu.default_stream() + time_cost = 0 + + while warm_ctr: _ = torchaie_model(input_tensor, input_tensor2) - print("finish infer once") - loops = 100 - warm_ctr = 10 - - default_stream = torch_aie.npu.default_stream() - time_cost = 0 - - while warm_ctr: - _ = torchaie_model(input_tensor, input_tensor2) - default_stream.synchronize() - warm_ctr -= 1 + default_stream.synchronize() + warm_ctr -= 1 - for i in range(loops): - t0 = time.time() - _ = torchaie_model(input_tensor, input_tensor2) - default_stream.synchronize() - t1 = time.time() - time_cost += (t1 - t0) - print(i) + for i in range(loops): + t0 = time.time() + _ = torchaie_model(input_tensor, input_tensor2) + default_stream.synchronize() + t1 = time.time() + time_cost += (t1 - t0) + print(i) - print(f"fps: {loops} * {batch_size} / {time_cost : .3f} samples/s") - print("torch_aie fps: ", loops * batch_size / time_cost) + print(f"fps: {loops} * {batch_size} / {time_cost : .3f} samples/s") + print("torch_aie fps: ", loops * batch_size / time_cost) from datetime import datetime current_time = datetime.now() @@ -89,4 +58,21 @@ if __name__ == '__main__': print("Current Time:", formatted_time) - +if __name__ == '__main__': + opts = parse_args() + batch_size = opts.batch_size + image_size = opts.image_size + ts_model = torch.jit.load(opts.ts_path) + input_info = [torch_aie.Input((batch_size, image_size)), torch_aie.Input((batch_size, 0))] + torch_aie.set_device(0) + print("start compile") + torchaie_model = torch_aie.compile( + ts_model, + inputs=input_info, + precision_policy=_enums.PrecisionPolicy.FP32, + soc_version='Ascend310P3', + ) + print("end compile") + torchaie_model.eval() + + perf(torchaie_model, batch_size, image_size) diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py index 0fbcaf6eff..cb4a06a604 100644 --- a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/perf_gpu.py @@ -1,4 +1,4 @@ -# Copyright 2021 Huawei Technologies Co., Ltd +# Copyright 2023 Huawei Technologies Co., Ltd # # Licensed under the Apache License, Version 2.0 (the License); # you may not use this file except in compliance with the License. diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_preprocess.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/preprocess.py similarity index 80% rename from AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_preprocess.py rename to AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/preprocess.py index cc2fd5cac9..1dcbbf8bcc 100644 --- a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/stylegan2-ada-pytorch_preprocess.py +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/preprocess.py @@ -1,4 +1,4 @@ -# Copyright 2021 Huawei Technologies Co., Ltd +# Copyright 2023 Huawei Technologies Co., Ltd # # Licensed under the Apache License, Version 2.0 (the License); # you may not use this file except in compliance with the License. @@ -52,6 +52,21 @@ import argparse from tqdm import tqdm +def parse_args(): + args = argparse.ArgumentParser(description="A program that generates model data to be predicted") + args.add_argument('--num_input', type=int, default=1) + args.add_argument('--batch_size', type=int, default=1) + args.add_argument('--save_path', type=str, default='./input') + return args.parse_args() + + +def check_args(args): + if args.num_input <= 0: + raise ValueError(f"num_input must be greater than 0. Got: {args.num_input}.") + if args.batch_size <= 0: + raise ValueError(f"batch_size must be greater than 0. Got: {args.batch_size}.") + + def main(args): # set up option z_dim = 512 @@ -71,12 +86,7 @@ def main(args): input.tofile(os.path.join(save_path, f'input_bs{bs}_{i:04d}.bin')) -# ---------------------------------------------------------------------------- if __name__ == "__main__": - parser = argparse.ArgumentParser() - parser.add_argument('--num_input', type=int, default=1) - parser.add_argument('--batch_size', type=int, default=1) - parser.add_argument('--save_path', type=str, default='./input') - args = parser.parse_args() + args = parse_args() + check_args(args) main(args) -# ---------------------------------------------------------------------------- diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_export.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_export.py new file mode 100644 index 0000000000..c5b7352dce --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_export.py @@ -0,0 +1,119 @@ +# Copyright 2023 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the License); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# BSD 3-Clause License +# +# Copyright (c) 2017 xxxx +# All rights reserved. +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# +# * Neither the name of the copyright holder nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# ============================================================================ + +import unittest +import argparse +import os +import shutil +from unittest.mock import patch + +from export import parse_args, check_args, trace_ts_model + + +class TestParseArgs(unittest.TestCase): + @patch('argparse.ArgumentParser.parse_args', return_value=argparse.Namespace( + stylegan_path='./stylegan', + pkl_file='model.pth' + )) + def test_parse_args_with_arguments(self, mock_parse_args): + # Call the function + args = parse_args() + + # Check if the function returns the expected arguments + self.assertEqual(args.stylegan_path, './stylegan') + self.assertEqual(args.pkl_file, 'model.pth') + + # Verify that argparse.ArgumentParser.parse_args was called + mock_parse_args.assert_called_once() + + +class TestCheckArgs(unittest.TestCase): + def test_nonexistent_stylegan_path(self): + args = argparse.Namespace(stylegan_path='nonexistent_stylegan.pth', pkl_file='nonexistent_model.pth') + + with self.assertRaises(FileNotFoundError) as context: + check_args(args) + expected_message = f'The stylegan2-ada-pytorch Github directory {args.stylegan_path} not exists' + self.assertEqual(str(context.exception), expected_message) + + def test_nonexistent_model_path(self): + stylegan_path = 'existing_stylegan' + os.makedirs(stylegan_path, exist_ok=True) + + args = argparse.Namespace(stylegan_path=stylegan_path, pkl_file='nonexistent_model.pth') + + with self.assertRaises(FileNotFoundError) as context: + check_args(args) + expected_message = f'The model file stylegan2-ada pth {args.pkl_file} not exists' + shutil.rmtree(stylegan_path) + self.assertEqual(str(context.exception), expected_message) + + +class TestTraceTSModel(unittest.TestCase): + def setUp(self): + pass + # self.temp_dir = 'temp_dir' + # os.makedirs(self.temp_dir, exist_ok=True) + + def tearDown(self): + pass + # shutil.rmtree(self.temp_dir) + + @patch('argparse.ArgumentParser.parse_args', return_value=argparse.Namespace( + stylegan_path='./stylegan2-ada-pytorch-main', + pkl_file='./models/G_ema_bs8_8p_kimg1000.pkl' + )) + def test_trace_ts_model(self, mock_parse_args): + args = parse_args() + trace_ts_model(args) + + # Check if the traced model file exists + self.assertTrue(os.path.exists("./stylegan2-bs.ts")) + + +if __name__ == '__main__': + unittest.main() diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_infer.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_infer.py new file mode 100644 index 0000000000..b15f7201db --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_infer.py @@ -0,0 +1,117 @@ +# Copyright 2023 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the License); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# BSD 3-Clause License +# +# Copyright (c) 2017 xxxx +# All rights reserved. +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# +# * Neither the name of the copyright holder nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# ============================================================================ + +import unittest +import argparse +import os +import shutil +from unittest.mock import patch + +from preprocess import parse_args, check_args, main +from infer import test_om + + +class TestParseArgs(unittest.TestCase): + @patch('argparse.ArgumentParser.parse_args', return_value=argparse.Namespace( + num_input=1, + batch_size=1 + )) + def test_parse_args_with_arguments(self, mock_parse_args): + # Call the function + args = parse_args() + + # Check if the function returns the expected arguments + self.assertEqual(args.num_input, 1) + self.assertEqual(args.batch_size, 1) + + # Verify that argparse.ArgumentParser.parse_args was called + mock_parse_args.assert_called_once() + + +class TestCheckArgs(unittest.TestCase): + def test_negative_num_input(self): + args = argparse.Namespace(num_input=-1, batch_size=1) + + with self.assertRaises(ValueError) as context: + check_args(args) + expected_message = f"num_input must be greater than 0. Got: {args.num_input}." + self.assertEqual(str(context.exception), expected_message) + + def test_negative_batch_size(self): + args = argparse.Namespace(num_input=1, batch_size=-1) + + with self.assertRaises(ValueError) as context: + check_args(args) + expected_message = f"batch_size must be greater than 0. Got: {args.batch_size}." + self.assertEqual(str(context.exception), expected_message) + + +class TestMain(unittest.TestCase): + def test_preprocess_main(self): + with patch('argparse.ArgumentParser.parse_args', return_value=argparse.Namespace( + num_input=200, + batch_size=1, + save_path="temp_path" + )): + args = parse_args() + check_args(args) + main(args) + files_in_folder = os.listdir(args.save_path) + self.assertEqual(len(files_in_folder), args.num_input, f"Expected {args.num_input} files, but found {len(files_in_folder)}.") + + def test_infer(self): + with patch('argparse.ArgumentParser.parse_args', return_value=argparse.Namespace( + bin_path="pre_data", # 假设已存在 + image_path="./results-bak", + ts_model_path="./stylegan2-bs.ts" + )): + args = parse_args() + test_om(args) + + +if __name__ == '__main__': + unittest.main() diff --git a/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_perf.py b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_perf.py new file mode 100644 index 0000000000..24404c8653 --- /dev/null +++ b/AscendIE/TorchAIE/built-in/cv/gan/stylegan2-ada/test_perf.py @@ -0,0 +1,91 @@ +# Copyright 2023 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the License); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# +# BSD 3-Clause License +# +# Copyright (c) 2017 xxxx +# All rights reserved. +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# +# * Neither the name of the copyright holder nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# ============================================================================ + +import unittest +import argparse +from unittest.mock import patch +import torch + +from perf import parse_args, perf + +import torch_aie +from torch_aie import _enums + + +class TestStyleGANPerf(unittest.TestCase): + def test_parse_args(self): + with patch('argparse.ArgumentParser.parse_args', return_value=argparse.Namespace( + ts_path='./stylegan2-bs.ts', + batch_size=1, + image_size=512 + )): + args = parse_args() + + self.assertEqual(args.ts_path, './stylegan2-bs.ts') + self.assertEqual(args.batch_size, 1) + self.assertEqual(args.image_size, 512) + + def test_perf(self): + with patch('argparse.ArgumentParser.parse_args', return_value=argparse.Namespace( + ts_path='./stylegan2-bs.ts', + batch_size=1, + image_size=512 + )): + args = parse_args() + ts_model = torch.jit.load(args.ts_path) + input_info = [torch_aie.Input((args.batch_size, args.image_size)), torch_aie.Input((args.batch_size, 0))] + torchaie_model = torch_aie.compile( + ts_model, + inputs=input_info, + precision_policy=_enums.PrecisionPolicy.FP32, # 必须是32才能跑 + soc_version='Ascend310P3' + ) + torchaie_model.eval() + perf(torchaie_model, args.batch_size, args.image_size) + + +if __name__ == '__main__': + unittest.main() -- Gitee