Merge branch 'fix/custom-models-default-nova2' into 'develop'

feat: default Custom Models fine-tuning to Nova 2 models

See merge request genaiic-reusable-assets/engagement-artifacts/genaiic-idp-accelerator!613

Author: rstrahan

config_library/finetuning_models.yaml

@@ -30,19 +30,10 @@ supported_models:
   - id: "us.amazon.nova-2-lite-v1:0"
     name: "Nova 2 Lite"
     provider: "Amazon"
-  
+
   - id: "us.amazon.nova-2-pro-v1:0"
     name: "Nova 2 Pro"
     provider: "Amazon"
-  
-  # Nova 1.x models (legacy)
-  - id: "us.amazon.nova-lite-v1:0"
-    name: "Nova Lite (v1)"
-    provider: "Amazon"
-  
-  - id: "us.amazon.nova-pro-v1:0"
-    name: "Nova Pro (v1)"
-    provider: "Amazon"
 
 # ---------------------------------------------------------------------------
 # Model ID Mappings for Fine-tuning API
@@ -55,14 +46,14 @@ supported_models:
 # Key: Base model ID (without cross-region prefix)
 # Value: Fine-tuning-capable model ID
 model_mappings:
-  # Nova 1.x mappings (300k context window)
-  "amazon.nova-lite-v1:0": "amazon.nova-lite-v1:0:300k"
-  "amazon.nova-pro-v1:0": "amazon.nova-pro-v1:0:300k"
-  
   # Nova 2.x mappings (256k context window)
   "amazon.nova-2-lite-v1:0": "amazon.nova-2-lite-v1:0:256k"
   "amazon.nova-2-pro-v1:0": "amazon.nova-2-pro-v1:0:256k"
 
+  # Nova 1.x mappings (legacy, 300k context window)
+  "amazon.nova-lite-v1:0": "amazon.nova-lite-v1:0:300k"
+  "amazon.nova-pro-v1:0": "amazon.nova-pro-v1:0:300k"
+
 # ---------------------------------------------------------------------------
 # Default Hyperparameters
 # ---------------------------------------------------------------------------
@@ -71,4 +62,4 @@ model_mappings:
 default_hyperparameters:
   epochCount: "2"
   learningRate: "0.00001"
-  batchSize: "1"
+  batchSize: "1"

lib/idp_common_pkg/idp_common/model_finetuning/service.py

@@ -28,9 +28,8 @@
 logger = logging.getLogger(__name__)
 
 # Default supported Nova models for fine-tuning (fallback if config not loaded)
+# Nova 2.x models are recommended; v1 models are kept for backward compatibility.
 DEFAULT_SUPPORTED_MODELS = [
-    {"id": "us.amazon.nova-lite-v1:0", "name": "Nova Lite", "provider": "Amazon"},
-    {"id": "us.amazon.nova-pro-v1:0", "name": "Nova Pro", "provider": "Amazon"},
     {"id": "us.amazon.nova-2-lite-v1:0", "name": "Nova 2 Lite", "provider": "Amazon"},
     {"id": "us.amazon.nova-2-pro-v1:0", "name": "Nova 2 Pro", "provider": "Amazon"},
 ]

lib/idp_common_pkg/tests/unit/test_model_finetuning.py

@@ -529,12 +529,10 @@ def test_list_available_models(self, service, mock_bedrock_client):
 
         result = service.list_available_models()
 
-        # Should have 4 base models (Nova Lite, Nova Pro, Nova 2 Lite, Nova 2 Pro)
-        assert len(result.base_models) == 4
-        assert result.base_models[0].name == "Nova Lite"
-        assert result.base_models[1].name == "Nova Pro"
-        assert result.base_models[2].name == "Nova 2 Lite"
-        assert result.base_models[3].name == "Nova 2 Pro"
+        # Should have 2 base models (Nova 2 Lite, Nova 2 Pro)
+        assert len(result.base_models) == 2
+        assert result.base_models[0].name == "Nova 2 Lite"
+        assert result.base_models[1].name == "Nova 2 Pro"
 
         # Should have 1 custom model
         assert len(result.custom_models) == 1

nested/appsync/src/lambda/finetuning_jobs_resolver/index.py

@@ -20,7 +20,6 @@
 from typing import Any, Dict, List, Optional
 
 import boto3
-from boto3.dynamodb.conditions import Key
 
 logger = logging.getLogger()
 logger.setLevel(logging.INFO)
@@ -39,10 +38,10 @@
 FINETUNING_JOB_PREFIX = "finetuning#"
 FINETUNING_JOBS_GSI_PK = "finetuning#jobs"
 
-# Supported base models for fine-tuning
+# Supported base models for fine-tuning (Nova 2.x recommended)
 SUPPORTED_BASE_MODELS = [
-    {"id": "us.amazon.nova-lite-v1:0", "name": "Nova Lite", "provider": "Amazon"},
-    {"id": "us.amazon.nova-pro-v1:0", "name": "Nova Pro", "provider": "Amazon"},
+    {"id": "us.amazon.nova-2-lite-v1:0", "name": "Nova 2 Lite", "provider": "Amazon"},
+    {"id": "us.amazon.nova-2-pro-v1:0", "name": "Nova 2 Pro", "provider": "Amazon"},
 ]
 
 
@@ -95,47 +94,56 @@ def lambda_handler(event: Dict[str, Any], context: Any) -> Any:
 
 def list_finetuning_jobs(arguments: Dict[str, Any]) -> Dict[str, Any]:
     """List all fine-tuning jobs with pagination.
-    
+
     Uses scan with filter since fine-tuning jobs are relatively few
     and the GSI1 index may not exist on all deployments.
     Jobs are stored with GSI1PK/GSI1SK for future GSI support.
+
+    The scan paginates through the table until enough matching items
+    are collected or the entire table has been scanned, because
+    DynamoDB's ``Limit`` caps items *evaluated* (before filtering),
+    not items *returned*.
     """
-    limit = arguments.get("limit", 20)
+    limit = arguments.get("limit", 50)
     next_token = arguments.get("nextToken")
 
     table = dynamodb.Table(TRACKING_TABLE_NAME)
 
-    # Use scan with filter - fine-tuning jobs have PK starting with 'finetuning#'
     from boto3.dynamodb.conditions import Attr
-    
-    scan_params = {
-        "FilterExpression": Attr("PK").begins_with(FINETUNING_JOB_PREFIX) & Attr("SK").eq("metadata"),
-        "Limit": limit * 5,  # Scan more items since filter reduces results
-    }
+
+    filter_expr = Attr("PK").begins_with(FINETUNING_JOB_PREFIX) & Attr("SK").eq(
+        "metadata"
+    )
+
+    # Collect all matching items by paginating through the scan.
+    # Fine-tuning jobs are few relative to the rest of the table,
+    # so a single scan page may not contain any matches.
+    items: List[Dict[str, Any]] = []
+    scan_kwargs: Dict[str, Any] = {"FilterExpression": filter_expr}
 
     if next_token:
-        scan_params["ExclusiveStartKey"] = json.loads(next_token)
+        scan_kwargs["ExclusiveStartKey"] = json.loads(next_token)
 
-    response = table.scan(**scan_params)
+    while True:
+        response = table.scan(**scan_kwargs)
 
-    # Filter and format items
-    items = []
-    for item in response.get("Items", []):
-        if item.get("PK", "").startswith(FINETUNING_JOB_PREFIX) and item.get("SK") == "metadata":
+        for item in response.get("Items", []):
             items.append(_format_job_for_graphql(item))
-    
+
+        # Stop if we've scanned the whole table
+        if "LastEvaluatedKey" not in response:
+            break
+
+        # Continue scanning from where we left off
+        scan_kwargs["ExclusiveStartKey"] = response["LastEvaluatedKey"]
+
     # Sort by createdAt descending (most recent first)
     items.sort(key=lambda x: x.get("createdAt", ""), reverse=True)
-    
+
     # Apply limit after sorting
     items = items[:limit]
 
-    result = {"items": items}
-
-    if "LastEvaluatedKey" in response and len(items) >= limit:
-        result["nextToken"] = json.dumps(response["LastEvaluatedKey"], cls=DecimalEncoder)
-
-    return result
+    return {"items": items}
 
 
 def get_finetuning_job(job_id: str) -> Optional[Dict[str, Any]]:

src/ui/src/components/custom-models/CustomModelsLayout.tsx

@@ -59,10 +59,10 @@ interface SelectOption {
   description?: string;
 }
 
-// Base models that support fine-tuning
+// Base models that support fine-tuning (Nova 2.x recommended)
 const SUPPORTED_BASE_MODELS: SelectOption[] = [
-  { label: 'Amazon Nova Pro', value: 'us.amazon.nova-pro-v1:0', description: 'High-performance model for complex tasks' },
-  { label: 'Amazon Nova Lite', value: 'us.amazon.nova-lite-v1:0', description: 'Balanced performance and cost' },
+  { label: 'Amazon Nova 2 Pro', value: 'us.amazon.nova-2-pro-v1:0', description: 'High-performance model for complex tasks' },
+  { label: 'Amazon Nova 2 Lite', value: 'us.amazon.nova-2-lite-v1:0', description: 'Balanced performance and cost' },
 ];
 
 // Status badge colors

src/ui/src/components/custom-models/FinetuningJobDetail.tsx

@@ -58,10 +58,12 @@ interface Notification {
   onDismiss: () => void;
 }
 
-// Base models that support fine-tuning
+// Base models that support fine-tuning (Nova 2.x + legacy v1 for display)
 const SUPPORTED_BASE_MODELS: Record<string, string> = {
-  'us.amazon.nova-pro-v1:0': 'Amazon Nova Pro',
-  'us.amazon.nova-lite-v1:0': 'Amazon Nova Lite',
+  'us.amazon.nova-2-pro-v1:0': 'Amazon Nova 2 Pro',
+  'us.amazon.nova-2-lite-v1:0': 'Amazon Nova 2 Lite',
+  'us.amazon.nova-pro-v1:0': 'Amazon Nova Pro (v1)',
+  'us.amazon.nova-lite-v1:0': 'Amazon Nova Lite (v1)',
 };
 
 // Status badge colors