新创建的AWS账号出于防恶意的角度,默认限制给的很低仅允许使用t3机型作为Notebook运行环境和推理节点。为进入正常业务使用,可能需要提升如下limit配额。
具体提升内容以实际业务使用为准。
SageMaker(Studio)
studio/JupyterServer-system 1
studio/KernelGateway-ml.t3.medium 2
studio/max_running_apps_per_domain 10
studio/max_user_profiles_per_domain 10
Notebook(Notebook Instance)
notebook-instance/ml.c4.2xlarge 1
notebook-instance/ml.c4.4xlarge 1
notebook-instance/ml.c5.2xlarge 1
notebook-instance/ml.c5.4xlarge 1
notebook-instance/ml.g4dn.2xlarge 1
notebook-instance/ml.g4dn.4xlarge 1
notebook-instance/ml.g4dn.8xlarge 1
notebook-instance/ml.t3.xlarge 1
训练(Training)
training-job/ml.p3.2xlarge 2
training-job/ml.p3.8xlarge 2
Batch Transform(批处理)
transform-job/ml.g4dn.2xlarge 2
transform-job/ml.g4dn.4xlarge 2
transform-job/ml.p2.xlarge 2
transform-job/ml.p3.2xlarge 2
transform-job/ml.p3.8xlarge 2
推理类(Endpoint)
endpoint/ml.c5.2xlarge 2
endpoint/ml.c5.4xlarge 2
endpoint/ml.g4dn.2xlarge 2
endpoint/ml.m5.2xlarge 2
endpoint/ml.p3.2xlarge 2
endpoint/total_accelerator_count 4
endpoint/total_instance_count 10
数据处理(Processing)
processing-job/ml.g4dn.2xlarge 2
processing-job/ml.g4dn.4xlarge 2
processing-job/ml.m5.2xlarge 2
processing-job/ml.m5.4xlarge 2
processing-job/ml.p3.2xlarge 2
processing-job/ml.p3.8xlarge 2
processing-job/total_instance_count 10
详情可参考如下文档:
https://docs.aws.amazon.com/general/latest/gr/sagemaker.html#limits_sagemaker