Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 22 additions & 11 deletions sagemaker-serve/src/sagemaker/serve/model_builder.py
Original file line number Diff line number Diff line change
Expand Up @@ -1015,25 +1015,36 @@ def _fetch_and_cache_recipe_config(self):
}

# Nova hosting configs per model (from Rhinestone modelDeployment.ts)
# NOTE: The nova-inference container (:SM-Inference-latest) enforces per-tier
# MAX_CONCURRENCY limits based on CONTEXT_LENGTH. These values were updated
# ~2026-03-23 synced with AGISageMakerInference ALLOWLISTED_CONFIGURATIONS.
# Uses the highest tier's CONTEXT_LENGTH and its MAX_CONCURRENCY per instance.
# If deployments fail with "MAX_CONCURRENCY N exceeds tier limit M", the
# container has likely tightened limits — check CloudWatch logs for the cap.
_NOVA_HOSTING_CONFIGS = {
"nova-textgeneration-micro": [
{"InstanceType": "ml.g5.12xlarge", "Environment": {"CONTEXT_LENGTH": "4096", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.g5.24xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "8192", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.g6.12xlarge", "Environment": {"CONTEXT_LENGTH": "10000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.g6.24xlarge", "Environment": {"CONTEXT_LENGTH": "10000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.g6.48xlarge", "Environment": {"CONTEXT_LENGTH": "12000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.p5.48xlarge", "Environment": {"CONTEXT_LENGTH": "12000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.g5.12xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "6"}},
{"InstanceType": "ml.g5.24xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "8"}},
{"InstanceType": "ml.g6.12xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "6"}},
{"InstanceType": "ml.g6.24xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "8"}},
{"InstanceType": "ml.g6.48xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "12"}},
{"InstanceType": "ml.g6e.xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "2"}},
{"InstanceType": "ml.g6e.2xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "2"}},
{"InstanceType": "ml.g6e.4xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "4"}},
{"InstanceType": "ml.p5.48xlarge", "Environment": {"CONTEXT_LENGTH": "128000", "MAX_CONCURRENCY": "8"}},
],
"nova-textgeneration-lite": [
{"InstanceType": "ml.g6.48xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "20000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.p5.48xlarge", "Environment": {"CONTEXT_LENGTH": "12000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.g6.12xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "2"}},
{"InstanceType": "ml.g6.24xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "4"}},
{"InstanceType": "ml.g6.48xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "8"}},
{"InstanceType": "ml.p5.48xlarge", "Environment": {"CONTEXT_LENGTH": "128000", "MAX_CONCURRENCY": "8"}},
],
"nova-textgeneration-pro": [
{"InstanceType": "ml.g6.48xlarge", "Environment": {"CONTEXT_LENGTH": "12000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.p5.48xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "50000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.p5.48xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "24000", "MAX_CONCURRENCY": "1"}},
],
"nova-textgeneration-lite-v2": [
{"InstanceType": "ml.p5.48xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "50000", "MAX_CONCURRENCY": "16"}},
{"InstanceType": "ml.g6.48xlarge", "Environment": {"CONTEXT_LENGTH": "8000", "MAX_CONCURRENCY": "8"}},
{"InstanceType": "ml.p5.48xlarge", "Profile": "Default", "Environment": {"CONTEXT_LENGTH": "128000", "MAX_CONCURRENCY": "8"}},
],
}

Expand Down
Loading