@aws/ml-container-creator 0.4.0 → 0.6.0

package/templates/do/config

@@ -10,6 +10,12 @@ export DEPLOYMENT_CONFIG="<%= deploymentConfig %>"
 export FRAMEWORK="<%= framework %>"
 export MODEL_SERVER="<%= modelServer %>"
 
+<% if (typeof enableLora !== 'undefined' && enableLora) { %>
+# LoRA adapter serving
+export ENABLE_LORA=true
+export ADAPTER_S3_BUCKET="mlcc-adapters-$(aws sts get-caller-identity --query Account --output text 2>/dev/null || echo 'UNKNOWN')-${AWS_REGION}"
+<% } %>
+
 # AWS configuration
 export AWS_REGION="<%= awsRegion %>"
 export ECR_REPOSITORY_NAME="ml-container-creator"
@@ -26,14 +32,27 @@ export DEPLOYMENT_TARGET="<%= deploymentTarget %>"
 
 <% if (deploymentTarget === 'realtime-inference') { %>
 # SageMaker Real-Time Inference configuration
+<% if (typeof existingEndpointName !== 'undefined' && existingEndpointName) { %>
+# External endpoint — attaching IC to an existing running endpoint
+export ENDPOINT_NAME="<%= existingEndpointName %>"
+export ENDPOINT_EXTERNAL=true
+<% } else { %>
 export INSTANCE_TYPE="<%= instanceType %>"
+<% if (typeof instancePools !== 'undefined' && instancePools && instancePools.length > 1) { %>
+# Instance pools: heterogeneous instance types with priority-based fallback
+# Priority = selection order (1 = preferred, higher = fallback)
+export INSTANCE_POOLS='<%= JSON.stringify(instancePools) %>'
+<% } %>
 <% if (inferenceAmiVersion) { %>
 export INFERENCE_AMI_VERSION="<%= inferenceAmiVersion %>"
 <% } %>
 <% if (typeof capacityReservationArn !== 'undefined' && capacityReservationArn) { %>
+# Note: Capacity reservations and instance pools (INSTANCE_POOLS) are mutually exclusive.
+# If both are set, the capacity reservation takes precedence and INSTANCE_POOLS is ignored.
 export CAPACITY_RESERVATION_ARN="<%= capacityReservationArn %>"
 <% } %>
 <% } %>
+<% } %>
 
 <% if (deploymentTarget === 'async-inference') { %>
 # SageMaker Async Inference configuration
@@ -49,7 +68,7 @@ ACCOUNT_ID=$(aws sts get-caller-identity --query Account --output text 2>/dev/nu
 <% if (asyncS3OutputPath) { %>
 export ASYNC_S3_OUTPUT_PATH="<%= asyncS3OutputPath %>"
 <% } else { %>
-export ASYNC_S3_OUTPUT_PATH="s3://ml-container-creator-async-${AWS_REGION}-${ACCOUNT_ID}/${PROJECT_NAME}/output/"
+export ASYNC_S3_OUTPUT_PATH="s3://mlcc-async-${ACCOUNT_ID}-${AWS_REGION}/${PROJECT_NAME}/output/"
 <% } %>
 
 <% if (asyncSnsSuccessTopic) { %>
@@ -89,12 +108,12 @@ ACCOUNT_ID=$(aws sts get-caller-identity --query Account --output text 2>/dev/nu
 <% if (batchInputPath) { %>
 export BATCH_INPUT_PATH="<%= batchInputPath %>"
 <% } else { %>
-export BATCH_INPUT_PATH="s3://ml-container-creator-batch-${AWS_REGION}-${ACCOUNT_ID}/${PROJECT_NAME}/input/"
+export BATCH_INPUT_PATH="s3://mlcc-batch-${ACCOUNT_ID}-${AWS_REGION}/${PROJECT_NAME}/input/"
 <% } %>
 <% if (batchOutputPath) { %>
 export BATCH_OUTPUT_PATH="<%= batchOutputPath %>"
 <% } else { %>
-export BATCH_OUTPUT_PATH="s3://ml-container-creator-batch-${AWS_REGION}-${ACCOUNT_ID}/${PROJECT_NAME}/output/"
+export BATCH_OUTPUT_PATH="s3://mlcc-batch-${ACCOUNT_ID}-${AWS_REGION}/${PROJECT_NAME}/output/"
 <% } %>
 export BATCH_INSTANCE_COUNT="<%= batchInstanceCount %>"
 export BATCH_SPLIT_TYPE="<%= batchSplitType %>"
@@ -205,7 +224,7 @@ export BENCHMARK_REQUEST_COUNT=""
 <% if (benchmarkS3OutputPath) { %>
 export BENCHMARK_S3_OUTPUT_PATH="<%= benchmarkS3OutputPath %>"
 <% } else { %>
-export BENCHMARK_S3_OUTPUT_PATH="s3://ml-container-creator-benchmark-${AWS_REGION}-$(aws sts get-caller-identity --query Account --output text)/${PROJECT_NAME}/"
+export BENCHMARK_S3_OUTPUT_PATH="s3://mlcc-benchmark-$(aws sts get-caller-identity --query Account --output text)-${AWS_REGION}/${PROJECT_NAME}/"
 <% } %>
 export BENCHMARK_JOB_NAME=""
 export BENCHMARK_WORKLOAD_CONFIG_NAME=""
@@ -222,7 +241,7 @@ export BASE_IMAGE=${BASE_IMAGE:-<%= baseImage || '' %>}
 
 # Allow environment variable overrides
 export AWS_REGION=${AWS_REGION:-<%= awsRegion %>}
-<% if (deploymentTarget === 'realtime-inference' || deploymentTarget === 'async-inference' || deploymentTarget === 'batch-transform') { %>
+<% if ((deploymentTarget === 'realtime-inference' && !(typeof existingEndpointName !== 'undefined' && existingEndpointName)) || deploymentTarget === 'async-inference' || deploymentTarget === 'batch-transform') { %>
 export INSTANCE_TYPE=${INSTANCE_TYPE:-<%= instanceType %>}
 <% } %>
 export ECR_REPOSITORY_NAME=${ECR_REPOSITORY_NAME:-ml-container-creator}
@@ -277,7 +296,11 @@ echo "   Model env vars: <%= Object.keys(modelEnvVars).length %>"
 echo "   Server env vars: <%= Object.keys(serverEnvVars).length %>"
 <% } %>
 <% if (deploymentTarget === 'realtime-inference') { %>
+<% if (typeof existingEndpointName !== 'undefined' && existingEndpointName) { %>
+echo "   Endpoint: ${ENDPOINT_NAME} (external)"
+<% } else { %>
 echo "   Instance: ${INSTANCE_TYPE}"
+<% } %>
 <% } else if (deploymentTarget === 'async-inference') { %>
 echo "   Instance: ${INSTANCE_TYPE}"
 echo "   S3 output: ${ASYNC_S3_OUTPUT_PATH}"