Spaces:
Runtime error
Runtime error
VenkateshRoshan
commited on
Commit
·
671ee28
1
Parent(s):
a562c0d
dockerfile updated
Browse files- dockerfile +0 -3
- src/deploy_sagemaker.py +122 -32
dockerfile
CHANGED
@@ -31,9 +31,6 @@ FROM python:3.10-slim
|
|
31 |
# # Run the application
|
32 |
# CMD ["python", "app.py"]
|
33 |
|
34 |
-
# Use NVIDIA CUDA base image
|
35 |
-
# FROM nvidia/cuda:12.1.0-runtime-ubuntu20.04
|
36 |
-
|
37 |
# Set environment variables
|
38 |
ENV PYTHONUNBUFFERED=TRUE
|
39 |
ENV PYTHONDONTWRITEBYTECODE=TRUE
|
|
|
31 |
# # Run the application
|
32 |
# CMD ["python", "app.py"]
|
33 |
|
|
|
|
|
|
|
34 |
# Set environment variables
|
35 |
ENV PYTHONUNBUFFERED=TRUE
|
36 |
ENV PYTHONDONTWRITEBYTECODE=TRUE
|
src/deploy_sagemaker.py
CHANGED
@@ -7,9 +7,40 @@ import os
|
|
7 |
from datetime import datetime
|
8 |
|
9 |
# Set up logging
|
10 |
-
logging.basicConfig(
|
|
|
|
|
|
|
11 |
logger = logging.getLogger(__name__)
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
def deploy_app(acc_id, region_name, role_arn, ecr_repo_name, endpoint_name="customer-support-chatbot"):
|
14 |
"""
|
15 |
Deploys a Gradio app as a SageMaker endpoint using an ECR image.
|
@@ -19,40 +50,99 @@ def deploy_app(acc_id, region_name, role_arn, ecr_repo_name, endpoint_name="cust
|
|
19 |
region_name (str): AWS region name
|
20 |
role_arn (str): IAM role ARN for SageMaker
|
21 |
ecr_repo_name (str): ECR repository name
|
22 |
-
endpoint_name (str): SageMaker endpoint name
|
23 |
"""
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
|
|
37 |
|
38 |
-
|
39 |
-
logger.info(f"Starting deployment of Gradio app to SageMaker endpoint {endpoint_name}...")
|
40 |
-
predictor = model.deploy(
|
41 |
-
initial_instance_count=1,
|
42 |
-
instance_type="ml.t3.large", #"ml.g4dn.xlarge",
|
43 |
-
endpoint_name=endpoint_name
|
44 |
-
)
|
45 |
-
logger.info(f"Gradio app deployed successfully to endpoint: {endpoint_name}")
|
46 |
-
|
47 |
-
if __name__ == "__main__":
|
48 |
-
# Parse arguments from CLI
|
49 |
parser = argparse.ArgumentParser(description="Deploy Gradio app to SageMaker")
|
50 |
-
parser.add_argument("--account_id", type=str, required=True,
|
51 |
-
|
52 |
-
parser.add_argument("--
|
53 |
-
|
54 |
-
parser.add_argument("--
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
args = parser.parse_args()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
|
57 |
-
|
58 |
-
|
|
|
7 |
from datetime import datetime
|
8 |
|
9 |
# Set up logging
|
10 |
+
logging.basicConfig(
|
11 |
+
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
|
12 |
+
level=logging.INFO
|
13 |
+
)
|
14 |
logger = logging.getLogger(__name__)
|
15 |
|
16 |
+
def create_model_archive(model_path):
|
17 |
+
"""
|
18 |
+
Create a model archive if needed
|
19 |
+
|
20 |
+
Args:
|
21 |
+
model_path (str): Path to model files
|
22 |
+
|
23 |
+
Returns:
|
24 |
+
str: S3 URI of the model archive
|
25 |
+
"""
|
26 |
+
try:
|
27 |
+
# Initialize S3 client
|
28 |
+
s3 = boto3.client('s3')
|
29 |
+
bucket = 'customer-support-gpt'
|
30 |
+
model_key = 'models/model.tar.gz'
|
31 |
+
|
32 |
+
# Check if model archive exists in S3
|
33 |
+
try:
|
34 |
+
s3.head_object(Bucket=bucket, Key=model_key)
|
35 |
+
logger.info("Model archive already exists in S3")
|
36 |
+
except:
|
37 |
+
logger.info("Model archive not found in S3, will be created during deployment")
|
38 |
+
|
39 |
+
return f's3://{bucket}/{model_key}'
|
40 |
+
except Exception as e:
|
41 |
+
logger.error(f"Error creating model archive: {str(e)}")
|
42 |
+
raise
|
43 |
+
|
44 |
def deploy_app(acc_id, region_name, role_arn, ecr_repo_name, endpoint_name="customer-support-chatbot"):
|
45 |
"""
|
46 |
Deploys a Gradio app as a SageMaker endpoint using an ECR image.
|
|
|
50 |
region_name (str): AWS region name
|
51 |
role_arn (str): IAM role ARN for SageMaker
|
52 |
ecr_repo_name (str): ECR repository name
|
53 |
+
endpoint_name (str): SageMaker endpoint name
|
54 |
"""
|
55 |
+
try:
|
56 |
+
logger.info("Starting SageMaker deployment process...")
|
57 |
+
|
58 |
+
# Initialize SageMaker session
|
59 |
+
sagemaker_session = sagemaker.Session()
|
60 |
+
|
61 |
+
# Define the image URI in ECR
|
62 |
+
ecr_image = f"{acc_id}.dkr.ecr.{region_name}.amazonaws.com/{ecr_repo_name}:latest"
|
63 |
+
logger.info(f"Using ECR image: {ecr_image}")
|
64 |
+
|
65 |
+
# Get model archive S3 URI
|
66 |
+
model_data = create_model_archive("models/customer_support_gpt")
|
67 |
+
|
68 |
+
# Define model configuration
|
69 |
+
model_environment = {
|
70 |
+
"MODEL_PATH": "/opt/ml/model",
|
71 |
+
"SAGEMAKER_SUBMIT_DIRECTORY": "/opt/ml/code",
|
72 |
+
"SAGEMAKER_PROGRAM": "inference.py"
|
73 |
+
}
|
74 |
+
|
75 |
+
# Create model
|
76 |
+
logger.info("Creating SageMaker model...")
|
77 |
+
model = Model(
|
78 |
+
image_uri=ecr_image,
|
79 |
+
model_data=model_data,
|
80 |
+
role=role_arn,
|
81 |
+
sagemaker_session=sagemaker_session,
|
82 |
+
env=model_environment,
|
83 |
+
enable_network_isolation=False
|
84 |
+
)
|
85 |
+
|
86 |
+
# Define deployment configuration
|
87 |
+
deployment_config = {
|
88 |
+
"initial_instance_count": 1,
|
89 |
+
"instance_type": "ml.t3.large",
|
90 |
+
"endpoint_name": endpoint_name,
|
91 |
+
"update_endpoint": True if _endpoint_exists(sagemaker_session, endpoint_name) else False
|
92 |
+
}
|
93 |
+
|
94 |
+
# Deploy model
|
95 |
+
logger.info(f"Deploying model to endpoint: {endpoint_name}")
|
96 |
+
logger.info(f"Deployment configuration: {deployment_config}")
|
97 |
+
|
98 |
+
predictor = model.deploy(**deployment_config)
|
99 |
+
|
100 |
+
logger.info(f"Successfully deployed to endpoint: {endpoint_name}")
|
101 |
+
return predictor
|
102 |
+
|
103 |
+
except Exception as e:
|
104 |
+
logger.error(f"Deployment failed: {str(e)}")
|
105 |
+
raise
|
106 |
|
107 |
+
def _endpoint_exists(sagemaker_session, endpoint_name):
|
108 |
+
"""Check if SageMaker endpoint already exists"""
|
109 |
+
client = sagemaker_session.boto_session.client('sagemaker')
|
110 |
+
try:
|
111 |
+
client.describe_endpoint(EndpointName=endpoint_name)
|
112 |
+
return True
|
113 |
+
except client.exceptions.ClientError:
|
114 |
+
return False
|
115 |
|
116 |
+
def main():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
117 |
parser = argparse.ArgumentParser(description="Deploy Gradio app to SageMaker")
|
118 |
+
parser.add_argument("--account_id", type=str, required=True,
|
119 |
+
help="AWS Account ID")
|
120 |
+
parser.add_argument("--region", type=str, required=True,
|
121 |
+
help="AWS Region")
|
122 |
+
parser.add_argument("--role_arn", type=str, required=True,
|
123 |
+
help="IAM Role ARN for SageMaker")
|
124 |
+
parser.add_argument("--ecr_repo_name", type=str, required=True,
|
125 |
+
help="ECR Repository name")
|
126 |
+
parser.add_argument("--endpoint_name", type=str,
|
127 |
+
default="customer-support-chatbot",
|
128 |
+
help="SageMaker Endpoint Name")
|
129 |
+
|
130 |
args = parser.parse_args()
|
131 |
+
|
132 |
+
try:
|
133 |
+
logger.info("Starting deployment process...")
|
134 |
+
deploy_app(
|
135 |
+
args.account_id,
|
136 |
+
args.region,
|
137 |
+
args.role_arn,
|
138 |
+
args.ecr_repo_name,
|
139 |
+
args.endpoint_name
|
140 |
+
)
|
141 |
+
logger.info("Deployment completed successfully!")
|
142 |
+
|
143 |
+
except Exception as e:
|
144 |
+
logger.error(f"Deployment failed: {str(e)}")
|
145 |
+
raise
|
146 |
|
147 |
+
if __name__ == "__main__":
|
148 |
+
main()
|