Error Using Local GPU Mode in SageMaker Notebook

Hello,

Currently receiving an error when fitting an estimator: I'm using an ml.p3.2xlarge. Instance type = "local_gpu"

from sagemaker.pytorch import PyTorch

env = {
    'SAGEMAKER_REQUIREMENTS': 'requirements.txt', # path relative to `source_dir` below.
}

estimator = PyTorch(
    entry_point="train.py",
    role=role,
    py_version="py3",
    framework_version= "1.7.1", #"1.8.0",
    instance_count=1,
    instance_type= instance_type
    hyperparameters={
        "n_classes": 3
    },
)

This is the error. I've been troubleshooting and even referenced Github official examples.

 Command '['docker', 'pull', '763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.7.1-gpu-py3']' returned non-zero exit status 1.

Any help would be appreciated.