Can I transfer vectors on disk to RAM?

[xtyDoge]

I have a collection with so many vectors, when I create it, I store the index on disk, see as below

{
  "result": {
    "status": "green",
    "optimizer_status": "ok",
    "vectors_count": 1100100,
    "indexed_vectors_count": 1100100,
    "points_count": 1100100,
    "segments_count": 8,
    "config": {
      "params": {
        "vectors": {
          "size": 512,
          "distance": "Dot",
          "on_disk": true
        },
        "shard_number": 2,
        "replication_factor": 1,
        "write_consistency_factor": 1,
        "on_disk_payload": true
      },
      "hnsw_config": {
        "m": 16,
        "ef_construct": 128,
        "full_scan_threshold": 10000,
        "max_indexing_threads": 4,
        "on_disk": false
      },
      "optimizer_config": {
        "deleted_threshold": 0.2,
        "vacuum_min_vector_number": 1000,
        "default_segment_number": 0,
        "max_segment_size": null,
        "memmap_threshold": null,
        "indexing_threshold": 10000,
        "flush_interval_sec": 5,
        "max_optimization_threads": 1
      },
      "wal_config": {
        "wal_capacity_mb": 32,
        "wal_segments_ahead": 0
      },
      "quantization_config": null
    },
    "payload_schema": {}
  },
  "status": "ok",
  "time": 0.000716049
}

Now I fount that I have enough RAM to load all of the vectors, so I changed the collection params like that

 PATCH /collections/t_feed_image_stress_test_202308
{
  "config": {
    "params": {
      "vectors": {
        "size": 512,
        "distance": "Euclid",
        "on_disk": false
      },
      "on_disk_payload": false
    }
  },
        "hnsw_config": {
        "m": 16,
        "ef_construct": 128,
        "full_scan_threshold": 10000,
        "max_indexing_threads": 4,
        "on_disk": false
      }
} 

But it not works，So how should I do to reach my goal?

[timvisee]

Could you elaborate on why it doesn't work? Do you receive an error, or is nothing happening?

You may have to explicitly specify the vector name, which will be an empty string for collections without named vectors. In that case, you should change your PATCH request to:

PATCH /collections/t_feed_image_stress_test_202308
{
  "config": {
    "params": {
      "vectors": {
        "": {
          "size": 512,
          "distance": "Euclid",
          "on_disk": false
        }
      },
      "on_disk_payload": false
    }
  },
  "hnsw_config": {
    "m": 16,
    "ef_construct": 128,
    "full_scan_threshold": 10000,
    "max_indexing_threads": 4,
    "on_disk": false
  }
} 

[xtyDoge]

After I use the patch method to update the collection params you offered, it returns "ok". And I try to use GET collections/t_feed_image_stress_test_202308 to get the detail of this collection after updating params. I expect the param
config.params.vectors."".on_disk and config.params.on_disk_payload should be false, but it still shows the old params before I execute the PATCH method. And the disk usage, RAM usage of the qdrant docker container also not changed. It seems like nothing happend.

{
  "result": {
    "status": "green",
    "optimizer_status": "ok",
    "vectors_count": 1100100,
    "indexed_vectors_count": 1100100,
    "points_count": 1100100,
    "segments_count": 8,
    "config": {
      "params": {
        "vectors": {
          "size": 512,
          "distance": "Dot",
          "on_disk": true
        },
        "shard_number": 2,
        "replication_factor": 1,
        "write_consistency_factor": 1,
        "on_disk_payload": true
      },
      "hnsw_config": {
        "m": 16,
        "ef_construct": 128,
        "full_scan_threshold": 10000,
        "max_indexing_threads": 4,
        "on_disk": false
      },
      "optimizer_config": {
        "deleted_threshold": 0.2,
        "vacuum_min_vector_number": 1000,
        "default_segment_number": 0,
        "max_segment_size": null,
        "memmap_threshold": null,
        "indexing_threshold": 10000,
        "flush_interval_sec": 5,
        "max_optimization_threads": 1
      },
      "wal_config": {
        "wal_capacity_mb": 32,
        "wal_segments_ahead": 0
      },
      "quantization_config": null
    },
    "payload_schema": {}
  },
  "status": "ok",
  "time": 0.000733874
}

[xtyDoge]

I successfully done with specify the vector name as "", I use the parms incorrect, so it doesn't work. Params below will work

PATCH /collections/t_feed_image_stress_test_202308
{
  "vectors": {
    "": {
      "on_disk": false
    }
  },
  "params": {
    "on_disk_payload": false
  }
}

Appreciate for your answer

[timvisee]

Fantastic! Happy to see you were able to fix it. Thank you for posting the answer.