feat: use different engine config to run sparse vector benchmarks

Author: KShivendu

.github/workflows/continuous-benchmark.yaml

@@ -17,7 +17,7 @@ jobs:
           ssh-private-key: ${{ secrets.SSH_PRIVATE_KEY }}
       - name: Benches
         run: |
-          DATASETS_TO_RUN=("laion-small-clip" "msmarco-sparse-100K")
+          DATASETS_TO_RUN=("laion-small-clip" "msmarco-sparse-1M")
           for dataset in "${DATASETS_TO_RUN[@]}"; do
             export DATASETS=$dataset
             export HCLOUD_TOKEN=${{ secrets.HCLOUD_TOKEN }}

experiments/configurations/qdrant-single-node.json

@@ -45,6 +45,26 @@
     ],
     "upload_params": { "parallel": 16, "batch_size": 1024 }
   },
+  {
+    "name": "qdrant-sparse-vector",
+    "engine": "qdrant",
+    "vector_type": "sparse",
+    "connection_params": { "timeout": 30 },
+    "collection_params": {
+      "optimizers_config": {
+        "max_segment_size": 1000000,
+        "default_segment_number": 3,
+        "memmap_threshold": 10000000
+      }
+    },
+    "search_params": [
+      {
+        "parallel": 8,
+        "search_params": {}
+      }
+    ],
+    "upload_params": { "parallel": 16, "batch_size": 1024 }
+  },
   {
     "name": "qdrant-parallel",
     "engine": "qdrant",