时间戳

首页 > 解决方案 > “NoneType”对象没有属性“boto_region_name”

python-3.x - “NoneType”对象没有属性“boto_region_name”

问题描述

我已经被这个问题缠住了一段时间，似乎无法理解这个问题来自哪里。我正在 Sagemaker 上部署模型，但在这行代码中出现错误：

sm_model.deploy(initial_instance_count=1, instance_type='ml.m4.2xlarge', endpoint_name=endpoint_name)

Jupyter Notebook 输出以下错误。注意：第 269 行不是我笔记本中的代码，它只是我通过上面的 model.deploy 命令获得的参考。

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)

    267             return self.image
    268 
--> 269         region_name = self.sagemaker_session.boto_region_name
    270         return create_image_uri(
    271             region_name,

AttributeError: 'NoneType' object has no attribute 'boto_region_name'

编辑：这只是我用来创建此管道的示例数据集。这是在 sagemaker 笔记本实例上。我在下面添加整个代码以进行澄清。

from sagemaker.model import Model
from sagemaker.pipeline import PipelineModel
import boto3
from time import gmtime, strftime

import boto3
import s3fs
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import seaborn as sns
%matplotlib inline
sns.set()



import boto3
import sagemaker
from sagemaker import get_execution_role

# Using Amazon S3
s3 = boto3.client('s3')
sage = boto3.client('sagemaker')

session = boto3.session.Session()    
sagemaker_session = sagemaker.Session()

# Get a SageMaker-compatible role used by this Notebook Instance.
role = get_execution_role()


#Upload file using AWS session
# S3 prefix

prefix = 'Scikit-keras-NLP-pipeline-Boston-Housing-example-June08-test1'


train_input = sagemaker_session.upload_data(
    path='housing.csv', 
    bucket=bucket,
    key_prefix='{}/{}'.format(prefix, 'train'))



from sagemaker.sklearn.estimator import SKLearn
output_dir = 's3://sagemaker-us-east-1-819182027957/Scikit-keras-NLP-pipeline-Boston-Housing-example-July08-test1/train'
model_dir = 's3://sagemaker-us-east-1-819182027957/Scikit-keras-NLP-pipeline-Boston-Housing-example-June08-test1/train'

script_path = 'Boston.py'

sklearn_preprocessor = SKLearn(
    entry_point=script_path,
    role=role,
    train_instance_type="ml.c4.xlarge",
    sagemaker_session=sagemaker_session,
    output_path=output_dir)


sklearn_preprocessor.fit({'train': train_input,'model-dir':model_dir,'output-data-dir':output_dir})






from sagemaker.tensorflow.serving import Model
sagemaker_estimator = Model(model_data = 's3://' + sagemaker_session.default_bucket() + '/Scikit-keras-NLP-pipeline-Boston-Housing-example-June08-test1/train/Bostonmodel.tar.gz',
                                  role = role)


#####


scikit_learn_inference_model = sklearn_preprocessor.create_model()
#sagemaker_model = sagemaker_estimator.create_model()                     # Does Not have create_model method
sagemaker_model = sagemaker_estimator




model_name = 'Boston-inf-pipeline-July08-model' 
endpoint_name = 'Boston-inf-pipeline-July08-endpoint'


#Build Inference Pipeline
sm_model = PipelineModel(
    name=model_name, 
    role=role, 
    models=[
        scikit_learn_inference_model, 
        sagemaker_model],
    sagemaker_session=sagemaker_session)

sm_model.deploy(initial_instance_count=1, instance_type='ml.c4.xlarge', endpoint_name=endpoint_name)

标签： python-3.xboto3botoamazon-sagemaker

解决方案

The error is caused by your sagemaker.tensorflow.serving.Model not having a sagemaker.session.Session associated with it.

Add sagemaker_session=sagemaker_session to your Model instantiation:

from sagemaker.tensorflow.serving import Model
sagemaker_model = Model(model_data='s3://' + sagemaker_session.default_bucket() + '/Scikit-keras-NLP-pipeline-Boston-Housing-example-June08-test1/train/Bostonmodel.tar.gz',
                        role=role,
                        sagemaker_session=sagemaker_session)

推荐阅读