Compute
Elastic Cloud Server
Huawei Cloud Flexus
Bare Metal Server
Auto Scaling
Image Management Service
Dedicated Host
FunctionGraph
Cloud Phone Host
Huawei Cloud EulerOS
Networking
Virtual Private Cloud
Elastic IP
Elastic Load Balance
NAT Gateway
Direct Connect
Virtual Private Network
VPC Endpoint
Cloud Connect
Enterprise Router
Enterprise Switch
Global Accelerator
Management & Governance
Cloud Eye
Identity and Access Management
Cloud Trace Service
Resource Formation Service
Tag Management Service
Log Tank Service
Config
OneAccess
Resource Access Manager
Simple Message Notification
Application Performance Management
Application Operations Management
Organizations
Optimization Advisor
IAM Identity Center
Cloud Operations Center
Resource Governance Center
Migration
Server Migration Service
Object Storage Migration Service
Cloud Data Migration
Migration Center
Cloud Ecosystem
KooGallery
Partner Center
User Support
My Account
Billing Center
Cost Center
Resource Center
Enterprise Management
Service Tickets
HUAWEI CLOUD (International) FAQs
ICP Filing
Support Plans
My Credentials
Customer Operation Capabilities
Partner Support Plans
Professional Services
Analytics
MapReduce Service
Data Lake Insight
CloudTable Service
Cloud Search Service
Data Lake Visualization
Data Ingestion Service
GaussDB(DWS)
DataArts Studio
Data Lake Factory
DataArts Lake Formation
IoT
IoT Device Access
Others
Product Pricing Details
System Permissions
Console Quick Start
Common FAQs
Instructions for Associating with a HUAWEI CLOUD Partner
Message Center
Security & Compliance
Security Technologies and Applications
Web Application Firewall
Host Security Service
Cloud Firewall
SecMaster
Anti-DDoS Service
Data Encryption Workshop
Database Security Service
Cloud Bastion Host
Data Security Center
Cloud Certificate Manager
Edge Security
Situation Awareness
Managed Threat Detection
Blockchain
Blockchain Service
Web3 Node Engine Service
Media Services
Media Processing Center
Video On Demand
Live
SparkRTC
MetaStudio
Storage
Object Storage Service
Elastic Volume Service
Cloud Backup and Recovery
Storage Disaster Recovery Service
Scalable File Service Turbo
Scalable File Service
Volume Backup Service
Cloud Server Backup Service
Data Express Service
Dedicated Distributed Storage Service
Containers
Cloud Container Engine
SoftWare Repository for Container
Application Service Mesh
Ubiquitous Cloud Native Service
Cloud Container Instance
Databases
Relational Database Service
Document Database Service
Data Admin Service
Data Replication Service
GeminiDB
GaussDB
Distributed Database Middleware
Database and Application Migration UGO
TaurusDB
Middleware
Distributed Cache Service
API Gateway
Distributed Message Service for Kafka
Distributed Message Service for RabbitMQ
Distributed Message Service for RocketMQ
Cloud Service Engine
Multi-Site High Availability Service
EventGrid
Dedicated Cloud
Dedicated Computing Cluster
Business Applications
Workspace
ROMA Connect
Message & SMS
Domain Name Service
Edge Data Center Management
Meeting
AI
Face Recognition Service
Graph Engine Service
Content Moderation
Image Recognition
Optical Character Recognition
ModelArts
ImageSearch
Conversational Bot Service
Speech Interaction Service
Huawei HiLens
Video Intelligent Analysis Service
Developer Tools
SDK Developer Guide
API Request Signing Guide
Terraform
Koo Command Line Interface
Content Delivery & Edge Computing
Content Delivery Network
Intelligent EdgeFabric
CloudPond
Intelligent EdgeCloud
Solutions
SAP Cloud
High Performance Computing
Developer Services
ServiceStage
CodeArts
CodeArts PerfTest
CodeArts Req
CodeArts Pipeline
CodeArts Build
CodeArts Deploy
CodeArts Artifact
CodeArts TestPlan
CodeArts Check
CodeArts Repo
Cloud Application Engine
MacroVerse aPaaS
KooMessage
KooPhone
KooDrive

Adding the Evaluation Code

Updated on 2022-12-08 GMT+08:00

After a training job is executed, ModelArts automatically evaluates your model and provides optimization diagnosis and suggestions. For details, see Viewing the Evaluation Result.

  • When you use a built-in algorithm to create a training job, you can view the evaluation result without any configurations.
  • For a training job created by compiling a training script or using a custom image, you need to add the evaluation code to the training code so that you can view the evaluation result and diagnosis suggestions after the training job is complete.
    NOTE:
    1. Only validation sets of the image type are supported.
    2. You can add the evaluation code only when the training scripts of the following frequently-used frameworks are used:
      • TF-1.13.1-python3.6
      • TF-2.1.0-python3.6
      • PyTorch-1.4.0-python3.6

This section describes how to use the evaluation code in a training job. To adapt and modify the training code, three steps are involved, adding the output path, copying the dataset to the local host, and mapping the dataset path to OBS.

Adding the Output Path

The code for adding the output path is simple. That is, add a path for storing the evaluation result file to the code, which is called train_url, that is, the training output path on the console. Add train_url to the analysis function and use save_path to obtain train_url. The sample code is as follows:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_string('model_url', '', 'path to saved model')
tf.app.flags.DEFINE_string('data_url', '', 'path to output files')
tf.app.flags.DEFINE_string('train_url', '', 'path to output files')
tf.app.flags.DEFINE_string('adv_param_json',
                           '{"attack_method":"FGSM","eps":40}',
                           'params for adversarial attacks')
FLAGS(sys.argv, known_only=True)

...

# analyse
res = analyse(
    task_type=task_type,
    pred_list=pred_list,
    label_list=label_list,
    name_list=file_name_list,
    label_map_dict=label_dict,
    save_path=FLAGS.train_url)

Copying the Dataset to the Local Host

Copying a dataset to the local host is to prevent the OBS connection from being interrupted due to long-time access. Therefore, copy the dataset to the local host before performing operations.

Datasets can be copied in two modes. Use the OBS path to copy datasets.

  • OBS path (recommended)

    Call the copy_parallel API of MoXing to copy the corresponding OBS path.

  • Dataset in ModelArts data management (manifest file format)

    Call the copy_manifest API of MoXing to copy the file to the local host and obtain the path of the new manifest file. Then, use SDK to parse the new manifest file.

1
2
3
4
5
6
7
8
if data_path.startswith('obs://'):
    if '.manifest' in data_path:
        new_manifest_path, _ = mox.file.copy_manifest(data_path, '/cache/data/')
        data_path = new_manifest_path
    else:
        mox.file.copy_parallel(data_path, '/cache/data/')
        data_path = '/cache/data/'
    print('------------- download dataset success ------------')

Mapping the Dataset Path to OBS

The actual path of the image file, that is, the OBS path, needs to be entered in the JSON body. Therefore, after analysis and evaluation are performed on the local host, the original local dataset path needs to be mapped to the OBS path, and the new list needs to be sent to the analysis interface.

If the OBS path is used as the input of data_url, you only need to replace the character string of the local path.

1
2
3
if FLAGS.data_url.startswith('obs://'):
    for idx, item in enumerate(file_name_list):
        file_name_list[idx] = item.replace(data_path, FLAGS.data_url)

If the manifest file is used, the original manifest file needs to be parsed again to obtain the list and then the list is sent to the analysis interface.

1
2
3
4
5
6
7
8
if or FLAGS.data_url.startswith('obs://'):
    if 'manifest' in FLAGS.data_url:
            file_name_list = []
            manifest, _ = get_sample_list(
                manifest_path=FLAGS.data_url, task_type='image_classification')
            for item in manifest:
                if len(item[1]) != 0:
                    file_name_list.append(item[0])

The sample code for image classification that adapts to a training job is as follows:

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
import json
import logging
import os
import sys
import tempfile

import h5py
import numpy as np
from PIL import Image

import moxing as mox
import tensorflow as tf
from deep_moxing.framework.manifest_api.manifest_api import get_sample_list
from deep_moxing.model_analysis.api import analyse, tmp_save
from deep_moxing.model_analysis.common.constant import TMP_FILE_NAME

logging.basicConfig(level=logging.DEBUG)

FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_string('model_url', '', 'path to saved model')
tf.app.flags.DEFINE_string('data_url', '', 'path to output files')
tf.app.flags.DEFINE_string('train_url', '', 'path to output files')
tf.app.flags.DEFINE_string('adv_param_json',
                           '{"attack_method":"FGSM","eps":40}',
                           'params for adversarial attacks')
FLAGS(sys.argv, known_only=True)


def _preprocess(data_path):
    img = Image.open(data_path)
    img = img.convert('RGB')
    img = np.asarray(img, dtype=np.float32)
    img = img[np.newaxis, :, :, :]
    return img


def softmax(x):
    x = np.array(x)
    orig_shape = x.shape
    if len(x.shape) > 1:
        # Matrix
        x = np.apply_along_axis(lambda x: np.exp(x - np.max(x)), 1, x)
        denominator = np.apply_along_axis(lambda x: 1.0 / np.sum(x), 1, x)
        if len(denominator.shape) == 1:
            denominator = denominator.reshape((denominator.shape[0], 1))
        x = x * denominator
    else:
        # Vector
        x_max = np.max(x)
        x = x - x_max
        numerator = np.exp(x)
        denominator = 1.0 / np.sum(numerator)
        x = numerator.dot(denominator)
    assert x.shape == orig_shape
    return x


def get_dataset(data_path, label_map_dict):
    label_list = []
    img_name_list = []
    if 'manifest' in data_path:
        manifest, _ = get_sample_list(
            manifest_path=data_path, task_type='image_classification')
        for item in manifest:
            if len(item[1]) != 0:
                label_list.append(label_map_dict.get(item[1][0]))
                img_name_list.append(item[0])
            else:
                continue
    else:
        label_name_list = os.listdir(data_path)
        label_dict = {}
        for idx, item in enumerate(label_name_list):
            label_dict[str(idx)] = item
            sub_img_list = os.listdir(os.path.join(data_path, item))
            img_name_list += [
                os.path.join(data_path, item, img_name) for img_name in sub_img_list
            ]
            label_list += [label_map_dict.get(item)] * len(sub_img_list)
    return img_name_list, label_list


def deal_ckpt_and_data_with_obs():
    pb_dir = FLAGS.model_url
    data_path = FLAGS.data_url

    if pb_dir.startswith('obs://'):
        mox.file.copy_parallel(pb_dir, '/cache/ckpt/')
        pb_dir = '/cache/ckpt'
        print('------------- download success ------------')
    if data_path.startswith('obs://'):
        if '.manifest' in data_path:
            new_manifest_path, _ = mox.file.copy_manifest(data_path, '/cache/data/')
            data_path = new_manifest_path
        else:
            mox.file.copy_parallel(data_path, '/cache/data/')
            data_path = '/cache/data/'
        print('------------- download dataset success ------------')
    assert os.path.isdir(pb_dir), 'Error, pb_dir must be a directory'
    return pb_dir, data_path


def evalution():
    pb_dir, data_path = deal_ckpt_and_data_with_obs()
    index_file = os.path.join(pb_dir, 'index')
    try:
        label_file = h5py.File(index_file, 'r')
        label_array = label_file['labels_list'][:].tolist()
        label_array = [item.decode('utf-8') for item in label_array]
    except Exception as e:
        logging.warning(e)
        logging.warning('index file is not a h5 file, try json.')
        with open(index_file, 'r') as load_f:
            label_file = json.load(load_f)
        label_array = label_file['labels_list'][:]
    label_map_dict = {}
    label_dict = {}
    for idx, item in enumerate(label_array):
        label_map_dict[item] = idx
        label_dict[idx] = item
    print(label_map_dict)
    print(label_dict)

    data_file_list, label_list = get_dataset(data_path, label_map_dict)

    assert len(label_list) > 0, 'missing valid data'
    assert None not in label_list, 'dataset and model not match'

    pred_list = []
    file_name_list = []
    img_list = []

    for img_path in data_file_list:
        img = _preprocess(img_path)
        img_list.append(img)
        file_name_list.append(img_path)

    config = tf.ConfigProto()
    config.gpu_options.allow_growth = True
    config.gpu_options.visible_device_list = '0'
    with tf.Session(graph=tf.Graph(), config=config) as sess:
        meta_graph_def = tf.saved_model.loader.load(
            sess, [tf.saved_model.tag_constants.SERVING], pb_dir)
        signature = meta_graph_def.signature_def
        signature_key = 'predict_object'
        input_key = 'images'
        output_key = 'logits'
        x_tensor_name = signature[signature_key].inputs[input_key].name
        y_tensor_name = signature[signature_key].outputs[output_key].name
        x = sess.graph.get_tensor_by_name(x_tensor_name)
        y = sess.graph.get_tensor_by_name(y_tensor_name)
        for img in img_list:
            pred_output = sess.run([y], {x: img})
            pred_output = softmax(pred_output[0])
            pred_list.append(pred_output[0].tolist())

    label_dict = json.dumps(label_dict)
    task_type = 'image_classification'

    if FLAGS.data_url.startswith('obs://'):
        if 'manifest' in FLAGS.data_url:
            file_name_list = []
            manifest, _ = get_sample_list(
                manifest_path=FLAGS.data_url, task_type='image_classification')
            for item in manifest:
                if len(item[1]) != 0:
                    file_name_list.append(item[0])
        for idx, item in enumerate(file_name_list):
            file_name_list[idx] = item.replace(data_path, FLAGS.data_url)
    # analyse
    res = analyse(
        task_type=task_type,
        pred_list=pred_list,
        label_list=label_list,
        name_list=file_name_list,
        label_map_dict=label_dict,
        save_path=FLAGS.train_url)

if __name__ == "__main__":
    evalution()

We use cookies to improve our site and your experience. By continuing to browse our site you accept our cookie policy. Find out more

Feedback

Feedback

Feedback

0/500

Selected Content

Submit selected content with the feedback