Create deploy-gptj.py

deploy-gptj.py

@@ -0,0 +1,25 @@
+
+from sagemaker.huggingface import HuggingFaceModel
+import sagemaker
+
+# IAM role with permissions to create endpoint
+role = sagemaker.get_execution_role()
+
+# public S3 URI to gpt-j artifact
+model_uri="s3://huggingface-sagemaker-models/transformers/4.12.3/pytorch/1.9.1/gpt-j/model.tar.gz"
+
+# create Hugging Face Model Class
+huggingface_model = HuggingFaceModel(
+    model_data=model_uri,
+    transformers_version='4.12.3',
+    pytorch_version='1.9.1',
+    py_version='py38',
+    role=role, 
+)
+
+# deploy model to SageMaker Inference
+predictor = huggingface_model.deploy(
+    initial_instance_count=1, # number of instances
+    instance_type='ml.g4dn.xlarge', #'ml.p3.2xlarge' # ec2 instance type
+    endpoint_name='sm-endpoint-gpt-j-6b'
+)