[西安交通大学][高校贡献][PyTorch离线推理][MAE]-初次提交-ModelZoo-PyTorch-AtomGit

AACL_PyTorch/contrib/cv/classfication/MAE/LICENSE+201-0

@@ -0,0 +1,201 @@
	1	+ Apache License
	2	+ Version 2.0, January 2004
	3	+ http://www.apache.org/licenses/
	4	+
	5	+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
	6	+
	7	+ 1. Definitions.
	8	+
	9	+ "License" shall mean the terms and conditions for use, reproduction,
	10	+ and distribution as defined by Sections 1 through 9 of this document.
	11	+
	12	+ "Licensor" shall mean the copyright owner or entity authorized by
	13	+ the copyright owner that is granting the License.
	14	+
	15	+ "Legal Entity" shall mean the union of the acting entity and all
	16	+ other entities that control, are controlled by, or are under common
	17	+ control with that entity. For the purposes of this definition,
	18	+ "control" means (i) the power, direct or indirect, to cause the
	19	+ direction or management of such entity, whether by contract or
	20	+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
	21	+ outstanding shares, or (iii) beneficial ownership of such entity.
	22	+
	23	+ "You" (or "Your") shall mean an individual or Legal Entity
	24	+ exercising permissions granted by this License.
	25	+
	26	+ "Source" form shall mean the preferred form for making modifications,
	27	+ including but not limited to software source code, documentation
	28	+ source, and configuration files.
	29	+
	30	+ "Object" form shall mean any form resulting from mechanical
	31	+ transformation or translation of a Source form, including but
	32	+ not limited to compiled object code, generated documentation,
	33	+ and conversions to other media types.
	34	+
	35	+ "Work" shall mean the work of authorship, whether in Source or
	36	+ Object form, made available under the License, as indicated by a
	37	+ copyright notice that is included in or attached to the work
	38	+ (an example is provided in the Appendix below).
	39	+
	40	+ "Derivative Works" shall mean any work, whether in Source or Object
	41	+ form, that is based on (or derived from) the Work and for which the
	42	+ editorial revisions, annotations, elaborations, or other modifications
	43	+ represent, as a whole, an original work of authorship. For the purposes
	44	+ of this License, Derivative Works shall not include works that remain
	45	+ separable from, or merely link (or bind by name) to the interfaces of,
	46	+ the Work and Derivative Works thereof.
	47	+
	48	+ "Contribution" shall mean any work of authorship, including
	49	+ the original version of the Work and any modifications or additions
	50	+ to that Work or Derivative Works thereof, that is intentionally
	51	+ submitted to Licensor for inclusion in the Work by the copyright owner
	52	+ or by an individual or Legal Entity authorized to submit on behalf of
	53	+ the copyright owner. For the purposes of this definition, "submitted"
	54	+ means any form of electronic, verbal, or written communication sent
	55	+ to the Licensor or its representatives, including but not limited to
	56	+ communication on electronic mailing lists, source code control systems,
	57	+ and issue tracking systems that are managed by, or on behalf of, the
	58	+ Licensor for the purpose of discussing and improving the Work, but
	59	+ excluding communication that is conspicuously marked or otherwise
	60	+ designated in writing by the copyright owner as "Not a Contribution."
	61	+
	62	+ "Contributor" shall mean Licensor and any individual or Legal Entity
	63	+ on behalf of whom a Contribution has been received by Licensor and
	64	+ subsequently incorporated within the Work.
	65	+
	66	+ 2. Grant of Copyright License. Subject to the terms and conditions of
	67	+ this License, each Contributor hereby grants to You a perpetual,
	68	+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
	69	+ copyright license to reproduce, prepare Derivative Works of,
	70	+ publicly display, publicly perform, sublicense, and distribute the
	71	+ Work and such Derivative Works in Source or Object form.
	72	+
	73	+ 3. Grant of Patent License. Subject to the terms and conditions of
	74	+ this License, each Contributor hereby grants to You a perpetual,
	75	+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
	76	+ (except as stated in this section) patent license to make, have made,
	77	+ use, offer to sell, sell, import, and otherwise transfer the Work,
	78	+ where such license applies only to those patent claims licensable
	79	+ by such Contributor that are necessarily infringed by their
	80	+ Contribution(s) alone or by combination of their Contribution(s)
	81	+ with the Work to which such Contribution(s) was submitted. If You
	82	+ institute patent litigation against any entity (including a
	83	+ cross-claim or counterclaim in a lawsuit) alleging that the Work
	84	+ or a Contribution incorporated within the Work constitutes direct
	85	+ or contributory patent infringement, then any patent licenses
	86	+ granted to You under this License for that Work shall terminate
	87	+ as of the date such litigation is filed.
	88	+
	89	+ 4. Redistribution. You may reproduce and distribute copies of the
	90	+ Work or Derivative Works thereof in any medium, with or without
	91	+ modifications, and in Source or Object form, provided that You
	92	+ meet the following conditions:
	93	+
	94	+ (a) You must give any other recipients of the Work or
	95	+ Derivative Works a copy of this License; and
	96	+
	97	+ (b) You must cause any modified files to carry prominent notices
	98	+ stating that You changed the files; and
	99	+
	100	+ (c) You must retain, in the Source form of any Derivative Works
	101	+ that You distribute, all copyright, patent, trademark, and
	102	+ attribution notices from the Source form of the Work,
	103	+ excluding those notices that do not pertain to any part of
	104	+ the Derivative Works; and
	105	+
	106	+ (d) If the Work includes a "NOTICE" text file as part of its
	107	+ distribution, then any Derivative Works that You distribute must
	108	+ include a readable copy of the attribution notices contained
	109	+ within such NOTICE file, excluding those notices that do not
	110	+ pertain to any part of the Derivative Works, in at least one
	111	+ of the following places: within a NOTICE text file distributed
	112	+ as part of the Derivative Works; within the Source form or
	113	+ documentation, if provided along with the Derivative Works; or,
	114	+ within a display generated by the Derivative Works, if and
	115	+ wherever such third-party notices normally appear. The contents
	116	+ of the NOTICE file are for informational purposes only and
	117	+ do not modify the License. You may add Your own attribution
	118	+ notices within Derivative Works that You distribute, alongside
	119	+ or as an addendum to the NOTICE text from the Work, provided
	120	+ that such additional attribution notices cannot be construed
	121	+ as modifying the License.
	122	+
	123	+ You may add Your own copyright statement to Your modifications and
	124	+ may provide additional or different license terms and conditions
	125	+ for use, reproduction, or distribution of Your modifications, or
	126	+ for any such Derivative Works as a whole, provided Your use,
	127	+ reproduction, and distribution of the Work otherwise complies with
	128	+ the conditions stated in this License.
	129	+
	130	+ 5. Submission of Contributions. Unless You explicitly state otherwise,
	131	+ any Contribution intentionally submitted for inclusion in the Work
	132	+ by You to the Licensor shall be under the terms and conditions of
	133	+ this License, without any additional terms or conditions.
	134	+ Notwithstanding the above, nothing herein shall supersede or modify
	135	+ the terms of any separate license agreement you may have executed
	136	+ with Licensor regarding such Contributions.
	137	+
	138	+ 6. Trademarks. This License does not grant permission to use the trade
	139	+ names, trademarks, service marks, or product names of the Licensor,
	140	+ except as required for reasonable and customary use in describing the
	141	+ origin of the Work and reproducing the content of the NOTICE file.
	142	+
	143	+ 7. Disclaimer of Warranty. Unless required by applicable law or
	144	+ agreed to in writing, Licensor provides the Work (and each
	145	+ Contributor provides its Contributions) on an "AS IS" BASIS,
	146	+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
	147	+ implied, including, without limitation, any warranties or conditions
	148	+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
	149	+ PARTICULAR PURPOSE. You are solely responsible for determining the
	150	+ appropriateness of using or redistributing the Work and assume any
	151	+ risks associated with Your exercise of permissions under this License.
	152	+
	153	+ 8. Limitation of Liability. In no event and under no legal theory,
	154	+ whether in tort (including negligence), contract, or otherwise,
	155	+ unless required by applicable law (such as deliberate and grossly
	156	+ negligent acts) or agreed to in writing, shall any Contributor be
	157	+ liable to You for damages, including any direct, indirect, special,
	158	+ incidental, or consequential damages of any character arising as a
	159	+ result of this License or out of the use or inability to use the
	160	+ Work (including but not limited to damages for loss of goodwill,
	161	+ work stoppage, computer failure or malfunction, or any and all
	162	+ other commercial damages or losses), even if such Contributor
	163	+ has been advised of the possibility of such damages.
	164	+
	165	+ 9. Accepting Warranty or Additional Liability. While redistributing
	166	+ the Work or Derivative Works thereof, You may choose to offer,
	167	+ and charge a fee for, acceptance of support, warranty, indemnity,
	168	+ or other liability obligations and/or rights consistent with this
	169	+ License. However, in accepting such obligations, You may act only
	170	+ on Your own behalf and on Your sole responsibility, not on behalf
	171	+ of any other Contributor, and only if You agree to indemnify,
	172	+ defend, and hold each Contributor harmless for any liability
	173	+ incurred by, or claims asserted against, such Contributor by reason
	174	+ of your accepting any such warranty or additional liability.
	175	+
	176	+ END OF TERMS AND CONDITIONS
	177	+
	178	+ APPENDIX: How to apply the Apache License to your work.
	179	+
	180	+ To apply the Apache License to your work, attach the following
	181	+ boilerplate notice, with the fields enclosed by brackets "[]"
	182	+ replaced with your own identifying information. (Don't include
	183	+ the brackets!) The text should be enclosed in the appropriate
	184	+ comment syntax for the file format. We also recommend that a
	185	+ file or class name and description of purpose be included on the
	186	+ same "printed page" as the copyright notice for easier
	187	+ identification within third-party archives.
	188	+
	189	+ Copyright [yyyy] [name of copyright owner]
	190	+
	191	+ Licensed under the Apache License, Version 2.0 (the "License");
	192	+ you may not use this file except in compliance with the License.
	193	+ You may obtain a copy of the License at
	194	+
	195	+ http://www.apache.org/licenses/LICENSE-2.0
	196	+
	197	+ Unless required by applicable law or agreed to in writing, software
	198	+ distributed under the License is distributed on an "AS IS" BASIS,
	199	+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	200	+ See the License for the specific language governing permissions and
	201	+ limitations under the License.

AACL_PyTorch/contrib/cv/classfication/MAE/MAE_postprocess.py+175-0

@@ -0,0 +1,175 @@
	1	+# Copyright 2021 Huawei Technologies Co., Ltd
	2	+#
	3	+# Licensed under the Apache License, Version 2.0 (the "License");
	4	+# you may not use this file except in compliance with the License.
	5	+# You may obtain a copy of the License at
	6	+#
	7	+# http://www.apache.org/licenses/LICENSE-2.0
	8	+#
	9	+# Unless required by applicable law or agreed to in writing, software
	10	+# distributed under the License is distributed on an "AS IS" BASIS,
	11	+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	12	+# See the License for the specific language governing permissions and
	13	+# limitations under the License.
	14	+# ============================================================================
	15	+
	16	+import os
	17	+import sys
	18	+import json
	19	+import time
	20	+import argparse
	21	+import numpy as np
	22	+
	23	+
	24	+#============================================================================
	25	+# Variables
	26	+#============================================================================
	27	+np.set_printoptions(threshold=sys.maxsize)
	28	+
	29	+LABEL_FILE = "HiAI_label.json"
	30	+
	31	+
	32	+#============================================================================
	33	+# Functions
	34	+#============================================================================
	35	+def gen_file_name(img_name):
	36	+ full_name = img_name.split('/')[-1]
	37	+ index = full_name.rfind('.')
	38	+ return full_name[:index]
	39	+
	40	+
	41	+def cre_groundtruth_dict(gtfile_path):
	42	+ img_gt_dict = {}
	43	+ for gtfile in os.listdir(gtfile_path):
	44	+ if (gtfile != LABEL_FILE):
	45	+ with open(os.path.join(gtfile_path, gtfile), 'r') as f:
	46	+ gt = json.load(f)
	47	+ ret = gt["image"]["annotations"][0]["category_id"]
	48	+ img_gt_dict[gen_file_name(gtfile)] = ret
	49	+ return img_gt_dict
	50	+
	51	+
	52	+def cre_groundtruth_dict_fromtxt(gtfile_path):
	53	+ img_count = 0
	54	+ img_gt_dict = {}
	55	+ with open(gtfile_path, 'r')as f:
	56	+ for line in f.readlines():
	57	+ img_count += 1
	58	+ temp = line.strip().split(" ")
	59	+ imgName = temp[0].split(".")[0]
	60	+ imgLab = temp[1]
	61	+ img_gt_dict[imgName] = imgLab
	62	+ return img_gt_dict, img_count
	63	+
	64	+
	65	+def load_statistical_predict_result(filepath, index):
	66	+ """
	67	+ the prediction esult file data extraction
	68	+ """
	69	+ with open(filepath, 'r')as f:
	70	+ for i, index_data in enumerate(f):
	71	+ if i == index:
	72	+ data = index_data
	73	+ temp = data.strip().split(" ")
	74	+ n_label = len(temp)
	75	+ if data == '':
	76	+ n_label = 0
	77	+ data_vec = np.zeros((n_label), dtype=np.float32)
	78	+ in_type = ''
	79	+ color = ''
	80	+ if n_label == 0:
	81	+ in_type = f.readline()
	82	+ color = f.readline()
	83	+ else:
	84	+ for ind, prob in enumerate(temp):
	85	+ data_vec[ind] = np.float32(prob)
	86	+ return data_vec, n_label, in_type, color
	87	+
	88	+
	89	+def create_visualization_statistical_result(prediction_file_path,
	90	+ result_store_path, json_file_name,
	91	+ batch_size, img_gt_dict,
	92	+ img_num, topn=5):
	93	+ writer = open(os.path.join(result_store_path, json_file_name), 'w')
	94	+ table_dict = {}
	95	+ table_dict["title"] = "Overall statistical evaluation"
	96	+ table_dict["value"] = []
	97	+
	98	+ count = 0
	99	+ resCnt = 0
	100	+ n_labels = 0
	101	+ count_hit = np.zeros(topn)
	102	+ for tfile_name in os.listdir(prediction_file_path):
	103	+ for i in range(batch_size):
	104	+ count += 1
	105	+ temp = tfile_name.split('.')[0]
	106	+ index = temp.find('_') + 1
	107	+ img_index = temp[index: index+5]
	108	+ convert_index = int(img_index) * batch_size + i + 1
	109	+ if convert_index > img_num:
	110	+ break
	111	+ img_name = "ILSVRC2012_val_{:08d}".format(convert_index)
	112	+ filepath = os.path.join(prediction_file_path, tfile_name)
	113	+ ret = load_statistical_predict_result(filepath, i)
	114	+ prediction = ret[0]
	115	+ n_labels = ret[1]
	116	+ sort_index = np.argsort(-prediction)
	117	+ gt = img_gt_dict[img_name]
	118	+ if (n_labels == 1000):
	119	+ realLabel = int(gt)
	120	+ elif (n_labels == 1001):
	121	+ realLabel = int(gt) + 1
	122	+ else:
	123	+ realLabel = int(gt)
	124	+
	125	+ resCnt = min(len(sort_index), topn)
	126	+ for j in range(resCnt):
	127	+ if (str(realLabel) == str(sort_index[j])):
	128	+ count_hit[j] += 1
	129	+ break
	130	+
	131	+ if 'value' not in table_dict.keys():
	132	+ print("the item value does not exist!")
	133	+ else:
	134	+ table_dict["value"].extend(
	135	+ [{"key": "Number of images", "value": str(count)},
	136	+ {"key": "Number of classes", "value": str(n_labels)}])
	137	+ if count == 0:
	138	+ accuracy = 0
	139	+ else:
	140	+ accuracy = np.cumsum(count_hit) / count
	141	+ for i in range(resCnt):
	142	+ table_dict["value"].append({"key": "Top" + str(i + 1) + " accuracy",
	143	+ "value": str(round(accuracy[i] * 100, 2)) + '%'})
	144	+ json.dump(table_dict, writer)
	145	+ writer.close()
	146	+
	147	+
	148	+#============================================================================
	149	+# Main
	150	+#============================================================================
	151	+if __name__ == '__main__':
	152	+ start = time.time()
	153	+
	154	+ parser = argparse.ArgumentParser()
	155	+ parser.add_argument('--folder-davinci-target', type=str, default="./result/outputs_bs1_om/")
	156	+ parser.add_argument('--annotation-file-path', type=str, default="/opt/npu/imageNet/val_label.txt")
	157	+ parser.add_argument('--result-json-path', type=str, default="./result")
	158	+ parser.add_argument('--json-file-name', type=str, default="result_bs1.json")
	159	+ parser.add_argument('--batch-size', type=int, default=1)
	160	+
	161	+ opt = parser.parse_args()
	162	+
	163	+ sub_folder = os.listdir(opt.folder_davinci_target)
	164	+ opt.folder_davinci_target = os.path.join(opt.folder_davinci_target, sub_folder[0])
	165	+
	166	+ img_label_dict, img_num = cre_groundtruth_dict_fromtxt(opt.annotation_file_path)
	167	+ create_visualization_statistical_result(opt.folder_davinci_target,
	168	+ opt.result_json_path,
	169	+ opt.json_file_name,
	170	+ opt.batch_size,
	171	+ img_label_dict,
	172	+ img_num, topn=1)
	173	+
	174	+ elapsed = (time.time() - start)
	175	+ print("Time used:", elapsed)

AACL_PyTorch/contrib/cv/classfication/MAE/MAE_preprocess.py+88-0

@@ -0,0 +1,88 @@
	1	+# Copyright 2021 Huawei Technologies Co., Ltd
	2	+#
	3	+# Licensed under the Apache License, Version 2.0 (the "License");
	4	+# you may not use this file except in compliance with the License.
	5	+# You may obtain a copy of the License at
	6	+#
	7	+# http://www.apache.org/licenses/LICENSE-2.0
	8	+#
	9	+# Unless required by applicable law or agreed to in writing, software
	10	+# distributed under the License is distributed on an "AS IS" BASIS,
	11	+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	12	+# See the License for the specific language governing permissions and
	13	+# limitations under the License.
	14	+# ============================================================================
	15	+
	16	+import os
	17	+import math
	18	+import torch
	19	+import argparse
	20	+import numpy as np
	21	+from tqdm import tqdm
	22	+
	23	+from PIL import Image
	24	+from torchvision import transforms
	25	+
	26	+
	27	+#============================================================================
	28	+# Functions
	29	+#============================================================================
	30	+def preprocess(img):
	31	+ input_transform = transforms.Compose([
	32	+ transforms.Resize(256, Image.BICUBIC),
	33	+ transforms.CenterCrop(224),
	34	+ transforms.ToTensor(),
	35	+ transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
	36	+ ])
	37	+ return input_transform(img)
	38	+
	39	+
	40	+def img_preprocess(args):
	41	+ global input_tensor
	42	+ save_path = os.path.realpath(args.prep_image)
	43	+ if not os.path.exists(save_path):
	44	+ os.makedirs(save_path)
	45	+
	46	+ in_files = os.listdir(args.image_path)
	47	+ file_list = []
	48	+ if not os.path.isfile(os.path.join(args.image_path, in_files[0])):
	49	+ for sub_dir in in_files:
	50	+ image_path = os.path.join(args.image_path, sub_dir)
	51	+ sub_file_list = os.listdir(image_path)
	52	+ for file in sub_file_list:
	53	+ file_list.append(os.path.join(image_path, file))
	54	+ else:
	55	+ for file in in_files:
	56	+ file_list.append(os.path.join(args.image_path, file))
	57	+
	58	+ suffix_len = -5
	59	+ file_list.sort(key=lambda x:int(x[suffix_len-8:suffix_len]))
	60	+ for i in tqdm(range(int(np.ceil(len(file_list) / args.batch_size)))):
	61	+
	Liang Zhen2022年6月22日这里建议用tdqm库来代替print打印显示进度，可视化效果更好
	62	+ for idx in range(args.batch_size):
	63	+ file_index = i * args.batch_size + idx
	64	+ if file_index < len(file_list):
	65	+ file = file_list[file_index]
	66	+ input_image = Image.open(file).convert('RGB')
	67	+ image_tensor = preprocess(input_image).unsqueeze(0)
	68	+ else:
	69	+ image_tensor = torch.zeros([1,3,224,224])
	70	+
	71	+ input_tensor = image_tensor if idx == 0 \
	72	+ else torch.cat([input_tensor, image_tensor], dim=0)
	73	+
	74	+ img = np.array(input_tensor).astype(np.float32)
	75	+ img.tofile(os.path.join(save_path, "input_{:05d}.bin".format(i)))
	76	+
	77	+
	78	+#============================================================================
	79	+# Main
	80	+#============================================================================
	81	+if __name__ == "__main__":
	82	+ parser = argparse.ArgumentParser()
	83	+ parser.add_argument('--image-path', type=str, default="/opt/npu/imageNet/val")
	84	+ parser.add_argument('--prep-image', type=str, default="./prep_image_bs1")
	85	+ parser.add_argument('--batch-size', type=int, default=1)
	86	+ opt = parser.parse_args()
	87	+
	88	+ img_preprocess(opt)

AACL_PyTorch/contrib/cv/classfication/MAE/MAE_pth2onnx.py+46-0

@@ -0,0 +1,46 @@
	1	+
	2	+# Copyright 2021 Huawei Technologies Co., Ltd
	3	+#
	4	+# Licensed under the Apache License, Version 2.0 (the "License");
	5	+# you may not use this file except in compliance with the License.
	6	+# You may obtain a copy of the License at
	7	+#
	8	+# http://www.apache.org/licenses/LICENSE-2.0
	9	+#
	10	+# Unless required by applicable law or agreed to in writing, software
	11	+# distributed under the License is distributed on an "AS IS" BASIS,
	12	+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	13	+# See the License for the specific language governing permissions and
	14	+# limitations under the License.
	15	+# ============================================================================
	16	+
	17	+import torch
	18	+import argparse
	19	+import sys
	20	+from mae import models_vit
	21	+
	22	+def pth2onnx(pth_path,output_file):
	23	+ device = torch.device('cpu')
	24	+ model = models_vit.__dict__['vit_base_patch16'](
	25	+ num_classes=1000,
	26	+ drop_path_rate=0.1,
	27	+ global_pool=True,
	28	+ )
	29	+ model.to(device)
	30	+ checkpoint = torch.load(pth_path,map_location=torch.device("cpu"))
	31	+ model.load_state_dict(checkpoint['model'],False)
	32	+ model.eval()
	33	+ input_names = ["image"]
	34	+ output_names = ["output"]
	Liang Zhen2022年6月22日我认为这里没有必要引入新变量 `model_without_ddp`，可以复用 `model`。同学再确认一下。
	35	+ dynamic_axes = {'image': {0: '-1'}, 'class': {0: '-1'}}
	36	+ dummy_input = torch.randn(1, 3, 224, 224)
	37	+ torch.onnx.export(model, dummy_input, output_file, opset_version=11,
	38	+ input_names=input_names,dynamic_axes=dynamic_axes, output_names=output_names, verbose=False)
	39	+
	Liang Zhen2022年6月22日单行代码不建议超过80字符；上库release代码建议设置 `verbose=Flase`
	40	+if __name__ == '__main__':
	41	+
	42	+ parser = argparse.ArgumentParser()
	43	+ parser.add_argument('--source', type=str, default="./mae_finetuned_vit_base.pth")
	44	+ parser.add_argument('--target', type=str, default="./mae_dynamicbs.onnx")
	45	+ args = parser.parse_args()
	46	+ pth2onnx(args.source,args.target)

AACL_PyTorch/contrib/cv/classfication/MAE/README.md+200-0

@@ -0,0 +1,200 @@
	1	+# MAE ONNX模型端到端推理指导
	2	+- [1. 模型概述](#1)
	3	+ - [论文地址](#11)
	4	+ - [代码地址](#12)
	5	+- [2. 环境说明](#2)
	6	+ - [深度学习框架](#21)
	7	+ - [python第三方库](#22)
	8	+- [3. 模型转换](#3)
	9	+ - [pth转onnx模型](#31)
	10	+ - [onnx转om模型](#32)
	11	+- [4. 数据预处理](#4)
	12	+ - [数据处理](#41)
	13	+- [5. 离线推理](#5)
	14	+ - [msame工具概述](#51)
	15	+ - [离线推理](#52)
	16	+ - [精度和性能比较](#53)
	17	+
	18	+## <a name="1">1. 模型概述</a>
	19	+### <a name="11">1.1 论文地址</a>
	20	+[MAE论文]([https://arxiv.org/pdf/2111.06377.pdf])
	21	+
	22	+### <a name="12">1.2 代码地址</a>
	23	+[MAE代码](https://github.com/facebookresearch/mae)
	Liang Zhen2022年6月22日这里需要指定代码仓分支名和 `commit_id`。防止开源代码仓更新后导致本模型代码出错或者不可用
	24	+
	25	+branch=main
	26	+
	27	+commit_id= be47fef7a727943547afb0c670cf1b26034c3c89
	28	+
	29	+## <a name="2">2. 环境说明</a>
	30	+### <a name="21">2.1 深度学习框架与第三方库</a>
	31	+```
	32	+CANN 5.1.RC1
	33	+timm==0.5.4
	34	+torchvision==0.6.0
	35	+torch==1.8.1
	36	+onnx==1.11.0
	37	+Pillow==9.1.1
	38	+numpy==1.21.6
	39	+tqdm==4.64.0
	40	+```
	41	+
	42	+> 说明：
	43	+> pytorch，torchvision和onnx:(X86架构)可以通过pip方式安装或官方下载whl包安装; (Arm架构)可以通过源码编译安装
	44	+> 其他第三方库: 可以通过 pip3 install -r requirements.txt 进行安装
	45	+
	46	+### <a name="21">2.2 权重文件</a>
	47	+
	48	+使用源仓提供的mae_finetuned_vit_base.pth
	49	+
	Liang Zhen2022年6月22日不建议直接给出链接，可参考如下描述使用开源仓提供的 mae_finetuned_vit_base.pth权重(提取码：1234)
	50	+链接：https://pan.baidu.com/s/1FwIK2db5nojOT7YC6rI1Hg
	51	+提取码：1234
	52	+
	53	+## <a name="3">3. 模型转换</a>
	54	+一步式从pth权重文件转om模型的脚本，能够由pth权重文件生成动态Batch的onnx模型和bacth为{batch_size}的om模型，{chip_name}为Ascend版本，可通过`npu-smi info`指令查看：
	55	+```bash
	56	+bash ./test/pth2om.sh --batch_size={batch_size} --not_skip_onnx=true {chip_name}
	57	+```
	58	+运行后会生成如下文件：
	59	+```bash
	60	+├── mae_dynamicbs.onnx
	61	+├── mae_bs{batch_size}.om
	62	+```
	63	+
	64	+### <a name="31">3.1 pth转onnx模型</a>
	65	+1. 设置环境变量
	66	+```bash
	67	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	68	+```
	Liang Zhen2022年6月22日内容不用重复，保留一处权重获取方式即可
	69	+
	70	+2. 执行MAE_pth2onnx.py脚本，生成onnx模型文件
	71	+
	72	+```bash
	73	+python3 MAE_pth2onnx.py --source "./mae_finetuned_vit_base.pth" --target "./mae_dynamicbs.onnx"
	74	+```
	75	+其中"source"表示模型加载权重的地址和名称，"target"表示转换后生成的onnx模型的存储地址和名称
	76	+
	77	+### <a name="32">3.2 onnx转om模型</a>
	78	+1. 使用atc将onnx模型转换为om模型文件，工具使用方法可以参考[CANN V100R020C10 开发辅助工具指南 (推理) 01](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/51RC2alpha002/infacldevg/atctool)
	79	+
	80	+ {batch_size}为batchsize，{chip_name}为Ascend版本，可通过`npu-smi info`指令查看
	81	+
	Liang Zhen2022年6月22日关于 `soc_version` 的设置建议直接参考复用 PR805 的相关描述
	82	+```bash
	83	+atc --framework=5 --model=mae_dynamicbs.onnx --output=mae_bs{batch_size} --input_format=NCHW --input_shape="image:{batch_size},3,224,224" --log=debug --soc_version={chip_name} --optypelist_for_implmode="Gelu" --op_select_implmode=high_performance --enable_small_channel=1
	84	+```
	85	+
	86	+## <a name="4">4. 数据预处理</a>
	87	+数据预处理过程包含在 test/eval_acc_perf.sh 的脚本中，可以直接运行，完成预处理+推理
	88	+### <a name="41">4.1 数据处理</a>
	89	+1. 设置环境变量
	90	+```bash
	91	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	92	+```
	93	+
	94	+2. MAE模型使用ImageNet2012中的5万张验证集数据进行测试，具体来说参考MAE的源码仓中的测试过程对验证集图像进行缩放，中心裁剪以及归一化，并将图像数据转换为二进制文件(.bin)
	95	+> 说明：
	96	+> 本项目使用的推理工具为msame，需要针对不同的batch size生成不同的输入数据
	97	+
	98	+3. 执行输入数据的生成脚本，生成模型输入的bin文件，以batch_size=1为例：
	99	+```bash
	100	+python3 MAE_preprocess.py --image-path /opt/npu/imageNet/val --prep-image ./prep_dataset_bs1/ --batch-size 1
	101	+```
	102	+其中"image-path"表示处理前原数据集的地址，"prep-image"表示生成数据集的文件夹名称(将在文件夹名称后会自动标识对应batch size，"batch-size"表示生成数据集对应的batch size
	103	+运行后，将会得到如下形式的文件夹：
	104	+
	105	+```
	106	+├── prep_dataset_bs1
	107	+│ ├──input_00000.bin
	108	+│ ├──......
	109	+```
	110	+
	111	+## <a name="5">5. 离线推理</a>
	112	+执行一步式推理前，请先准备msame离线推理工具
	113	+一步式进行输入数据的准备，模型离线推理和NPU性能数据的获取：
	114	+
	115	+```bash
	116	+bash ./test/eval_acc_perf.sh --batch_size=1 --datasets_path=/opt/npu/imageNet
	117	+```
	118	+运行后会生成如下文件/文件夹：
	119	+```bash
	120	+├── prep_dataset_bs1 # 模型的标签输入(文件夹)
	121	+├── msame_bs1.txt # msame推理过程的输出
	122	+├── result
	123	+│ ├── outputs_bs1_om # 模型的输出(文件夹)
	124	+│ ├── result_bs1.json # 模型的精度输出
	125	+```
	126	+
	127	+### <a name="51">5.1 msame工具概述</a>
	128	+msame模型推理工具，其输入是om模型以及模型所需要的输入bin文件，其输出是模型根据相应输入产生的输出文件。获取工具及使用方法可以参考[msame模型推理工具指南](https://gitee.com/ascend/tools/tree/master/msame)
	129	+### <a name="52">5.2 离线推理</a>
	130	+1. 设置环境变量
	131	+```bash
	132	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	133	+```
	134	+
	135	+2. 执行离线推理
	136	+ 运行如下命令进行离线推理：
	137	+
	138	+ Batch_size=1
	139	+```bash
	140	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	141	+mkdir result
	142	+./msame --model ./mae_bs1.om --output ./result/outputs_bs1 --outfmt TXT --input ./prep_dataset_bs1
	143	+```
	144	+ Batch_size=8
	145	+
	146	+```
	147	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	148	+mkdir result
	149	+./msame --model ./mae_bs8.om --output ./result/outputs_bs8 --outfmt TXT --input ./prep_dataset_bs8
	150	+```
	151	+
	152	+模型输出格式是txt，输出保存在"output"参数指定的文件夹中，同时会生成推理的日志文件msame_bs1.txt
	153	+
	154	+3.使用msame工具进行纯推理，可以根据不同的batchsize进行纯推理。
	155	+
	156	+```
	157	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	158	+mkdir result
	159	+./msame --model ./mae_bs{batch_size}.om --output ./result/outputs_bs{batch_size}_om --outfmt TXT --loop 20
	160	+```
	161	+
	162	+
	163	+
	164	+### <a name="53">5.3 精度和性能比较</a>
	165	+1. 性能数据的获取
	166	+通过给test/parser.py指定推理后的日志文件，可以得到离线推理的性能数据
	167	+```bash
	168	+python3 test/parse.py --result-file ./msame_bs1.txt --batch-size 1
	169	+```
	170	+其中"result-file"表示性能数据的地址和名称，"batch-size"表示性能测试时模型对应的batch size
	171	+
	172	+2. 精度数据的计算
	173	+精度计算利用MAE_postprocess.py脚本
	174	+```
	175	+python3 MAE_postprocess.py --folder-davinci-target ./result/outputs_bs1/ --annotation-file-path /opt/npu/imageNet/val_label.txt --result-json-path ./result --json-file-name result_bs1.json --batch-size 1
	176	+```
	177	+其中"folder-davinci-target"表示离线推理输出所在的文件夹，"annotation-file-path"表示ImageNet2012验证集标签的地址和名称，"result-json-path"输出精度数据所在的文件夹，"json-file-name"表示输出精度数据所在的文件名，"batch-size"表示精度测试时模型对应的batch size
	178	+
	179	+3. 精度数据的获取
	180	+通过给test/parser.py指定推理后的日志文件，可以得到离线推理的精度数据
	181	+```bash
	182	+python3 test/parse.py --result-file ./result/result_bs1.json
	183	+```
	184	+\| 模型 \| 参考精度 \| 310P精度 \| 性能基准 \| 310P性能 \|
	185	+\| :------: \| :------: \| :------: \| :------: \| :------: \|
	186	+\| mae bs1 \| top1:83.66% \| top1:83.5% \| 193.396fps \| 247.525fps \|
	187	+\| mae bs8 \| top1:83.66% \| top1:83.5% \| 272.124fps \| 450.988fps \|
	188	+
	189	+> 说明：
	190	+> Top1表示预测结果中概率最大的类别与真实类别一致的概率，其值越大说明分类模型的效果越优
	191	+
	192	+\| 模型 \| 性能基准 \| 310P性能 \|
	193	+\| :------: \| :--------: \| :--------: \|
	194	+\| mae bs1 \| 193.396fps \| 247.525fps \|
	195	+\| mae bs4 \| 257.646fps \| 283.168fps \|
	196	+\| mae bs8 \| 272.124fps \| 450.988fps \|
	197	+\| mae bs16 \| 263.026fps \| 320.32fps \|
	198	+\| mae bs32 \| 269.874fps \| 308.447fps \|
	199	+\| mae bs64 \| 266.179fps \| 296.744fps \|
	200	+

AACL_PyTorch/contrib/cv/classfication/MAE/modelzoo_level.txt+2-0

@@ -0,0 +1,2 @@
	1	+ModelConvert:OK
	2	+QuantStatus:OK

AACL_PyTorch/contrib/cv/classfication/MAE/requirements.txt+4-0

@@ -0,0 +1,4 @@
	1	+timm==0.5.4
	Liang Zhen2022年6月22日上面的 `msame` 二进制文件不用上传了
	2	+Pillow==9.1.1
	3	+numpy==1.21.6
	4	+tqdm==4.64.0

AACL_PyTorch/contrib/cv/classfication/MAE/test/eval_acc_perf.sh+79-0

@@ -0,0 +1,79 @@
	1	+#!/bin/bash
	2	+
	3	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	4	+
	5	+batch_size=1
	6	+datasets_path="/opt/npu/imageNet"
	7	+
	8	+for para in $*
	9	+do
	10	+ if [[ $para == --datasets_path* ]]; then
	11	+ datasets_path=`echo ${para#*=}`
	12	+ fi
	13	+ if [[ $para == --batch_size* ]]; then
	14	+ batch_size=`echo ${para#*=}`
	15	+ fi
	16	+done
	17	+
	18	+# ======================= generate prep_dataset ==============================
	19	+rm -rf ./prep_image_bs${batch_size}
	20	+python MAE_preprocess.py \
	21	+ --image-path ${datasets_path}/val \
	22	+ --prep-image ./prep_dataset_bs${batch_size} \
	23	+ --batch-size ${batch_size}
	24	+if [ $? != 0 ]; then
	25	+ echo "convmixer preprocess fail!"
	26	+ exit -1
	27	+fi
	28	+echo "==> 1. creating ./prep_image_bs${batch_size} successfully."
	29	+
	30	+# =============================== msame ======================================
	31	+
	32	+if [ ! -d ./result ]; then
	33	+ mkdir ./result
	34	+fi
	35	+rm -rf ./result/outputs_bs${batch_size}_om
	36	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	37	+echo "==> conducting. Please wait a moment!"
	38	+./msame --model ./mae_bs${batch_size}.om --output ./result/outputs_bs${batch_size}_om --outfmt TXT --input ./prep_dataset_bs${batch_size} > msame_bs${batch_size}.txt
	39	+if [ $? != 0 ]; then
	40	+ echo "msame bs${batch_size} fail!"
	41	+ exit -1
	42	+fi
	43	+echo "==> 2. conducting mae_bs${batch_size}.om successfully."
	44	+
	45	+
	46	+# ============================ evaluate ======================================
	47	+python3 MAE_postprocess.py \
	48	+ --folder-davinci-target ./result/outputs_bs${batch_size}_om/ \
	49	+ --annotation-file-path ${datasets_path}/val_label.txt \
	50	+ --result-json-path ./result \
	51	+ --json-file-name result_bs${batch_size}.json \
	52	+ --batch-size ${batch_size}
	53	+if [ $? != 0 ]; then
	54	+ echo "fail!"
	55	+ exit -1
	56	+fi
	57	+echo "==> 3. evaluating mae on bs${batch_size} successfully."
	58	+
	59	+
	60	+# =========================== print performance data =========================
	61	+echo "====performance data===="
	62	+python test/parse.py --result-file ./msame_bs${batch_size}.txt --batch-size ${batch_size}
	63	+if [ $? != 0 ]; then
	64	+ echo "parse bs${batch_size} performance fail!"
	65	+ exit -1
	66	+fi
	67	+
	68	+
	69	+# =========================== print accuracy data ============================
	70	+echo "====accuracy data===="
	71	+python test/parse.py --result-file ./result/result_bs${batch_size}.json
	72	+if [ $? != 0 ]; then
	73	+ echo "parse bs${batch_size} accuracy fail!"
	74	+ exit -1
	75	+fi
	76	+
	77	+
	78	+echo '==> 4. printing performance and accuracy data successfully.'
	79	+echo '==> 5. Done.'

AACL_PyTorch/contrib/cv/classfication/MAE/test/parse.py+40-0

@@ -0,0 +1,40 @@
	1	+# Copyright 2021 Huawei Technologies Co., Ltd
	2	+#
	3	+# Licensed under the Apache License, Version 2.0 (the "License");
	4	+# you may not use this file except in compliance with the License.
	5	+# You may obtain a copy of the License at
	6	+#
	7	+# http://www.apache.org/licenses/LICENSE-2.0
	8	+#
	9	+# Unless required by applicable law or agreed to in writing, software
	10	+# distributed under the License is distributed on an "AS IS" BASIS,
	11	+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	12	+# See the License for the specific language governing permissions and
	13	+# limitations under the License.
	14	+# ============================================================================
	15	+
	16	+import re
	17	+import sys
	18	+import json
	19	+import argparse
	20	+
	21	+if __name__ == '__main__':
	22	+ parser = argparse.ArgumentParser()
	23	+ parser.add_argument('--result-file', type=str, default="./msame_bs1.txt")
	24	+ parser.add_argument('--batch-size', type=int, default=1)
	25	+ args = parser.parse_args()
	26	+
	27	+ if args.result_file.endswith('.json'):
	28	+ result_json = args.result_file
	29	+ with open(result_json, 'r') as f:
	30	+ content = f.read()
	31	+ tops = [i.get('value') for i in json.loads(content).get('value') if 'Top' in i.get('key')]
	32	+ print('om {} top1:{}'.format(result_json.split('_')[1].split('.')[0], tops[0]))
	33	+ elif args.result_file.endswith('.txt'):
	34	+ result_txt = args.result_file
	35	+ with open(result_txt, 'r') as f:
	36	+ content = f.read()
	37	+ txt_data_list = re.findall(r'Inference average time without first time:.*ms', content.replace('\n', ',') + ',')[-1]
	38	+ avg_time = txt_data_list.split(' ')[-2]
	39	+ fps = args.batch_size * 1000 / float(avg_time)
	40	+ print('310P bs{} fps:{:.3f}'.format(args.batch_size, fps))

AACL_PyTorch/contrib/cv/classfication/MAE/test/perf_t4.sh+33-0

@@ -0,0 +1,33 @@
	1	+#!/bin/bash
	2	+
	3	+
	4	+rm -rf perf_bs1.log
	5	+trtexec --onnx=./mae_dynamicbs.onnx --shapes=image:1x3x224x224 --fp16 --threads > perf_bs1.log
	6	+perf_str=`grep "GPU.* mean.*ms$" perf_bs1.log`
	7	+if [ -n "$perf_str" ]; then
	8	+ perf_num=`echo $perf_str \| awk -F' ' '{print $16}'`
	9	+else
	10	+ perf_str=`grep "mean.*ms$" perf_bs1.log`
	11	+ perf_num=`echo $perf_str \| awk -F' ' '{print $4}'`
	12	+fi
	13	+echo "==> mae bs=1."
	14	+awk 'BEGIN{printf "t4 bs1 fps:%.3f\n", 1000*1/('$perf_num'/1)}'
	15	+
	16	+
	17	+
	18	+rm -rf perf_bs8.log
	19	+trtexec --onnx=./mae_dynamicbs.onnx --shapes=image:8x3x224x224 --fp16 --threads > perf_bs8.log
	20	+perf_str=`grep "GPU.* mean.*ms$" perf_bs4.log`
	21	+if [ -n "$perf_str" ]; then
	22	+ perf_num=`echo $perf_str \| awk -F' ' '{print $16}'`
	23	+else
	24	+ perf_str=`grep "mean.*ms$" perf_bs8.log`
	25	+ perf_num=`echo $perf_str \| awk -F' ' '{print $4}'`
	26	+fi
	27	+echo "==> mae bs=8."
	28	+awk 'BEGIN{printf "t4 bs8 fps:%.3f\n", 1000*1/('$perf_num'/1)}'
	29	+
	30	+
	31	+
	32	+
	33	+

AACL_PyTorch/contrib/cv/classfication/MAE/test/pth2om.sh+42-0

@@ -0,0 +1,42 @@
	1	+#!/bin/bash
	2	+
	3	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	4	+
	5	+batch_size=1
	6	+not_skip_onnx=true
	7	+chip_name==$3
	8	+
	9	+for para in $*
	10	+do
	11	+ if [[ $para == --batch_size* ]]; then
	12	+ batch_size=`echo ${para#*=}`
	13	+ fi
	14	+ if [[ $para == --not_skip_onnx* ]]; then
	15	+ not_skip_onnx=`echo ${para#*=}`
	16	+ fi
	17	+done
	18	+
	19	+# ======================= convert onnx =======================================
	20	+if [ $not_skip_onnx == true ]; then
	21	+ rm -rf mae_dynamicbs.onnx
	22	+ python MAE_pth2onnx.py \
	23	+ --source "./mae_finetuned_vit_base.pth" \
	24	+ --target "./mae_dynamicbs.onnx"
	25	+ if [ -f "./mae_dynamicbs.onnx" ]; then
	26	+ echo "==> 1. creating onnx model successfully."
	27	+ else
	28	+ echo "onnx export failed"
	29	+ exit -1
	30	+ fi
	31	+fi
	32	+
	33	+# ======================= convert om =========================================
	34	+rm -rf mae_bs${batch_size}.om
	35	+source /usr/local/Ascend/ascend-toolkit/set_env.sh
	36	+atc --framework=5 --model=mae_dynamicbs.onnx --output=mae_bs${batch_size} --input_format=NCHW --input_shape="image:${batch_size},3,224,224" --log=debug --soc_version=$3 --optypelist_for_implmode="Gelu" --op_select_implmode=high_performance --enable_small_channel=1
	37	+if [ -f "mae_bs${batch_size}.om" ] ; then
	38	+ echo "==> 2. creating om model successfully."
	39	+else
	40	+ echo "om export failed"
	41	+fi
	42	+echo "==> 3. Done."