1. Describe the bug
When running the eval_performance script, the reported memory_usage metric shows an impossibly large value: 16777216.00 TB. This is clearly a bug in how memory is being calculated or displayed.
2. To Reproduce
- Run the
eval_performance script with following configuration.
# PQ Performance Test Configuration
global:
exporters:
e1:
format: "text"
to: "stdout"
e2:
format: "json"
to: "file://tmp/github-perf.json"
SIFT-pq:
datapath: "/tmp/data/sift-128-euclidean.hdf5"
type: "build"
index_name: "ivf"
create_params: '{"dim":128,"dtype":"float32","metric_type":"l2","index_param": {
"buckets_count": 50,
"partition_strategy_type": "ivf",
"ivf_train_type": "kmeans",
"base_quantization_type": "pqfs",
"precise_quantization_type": "fp32",
"use_reorder": true,
"base_pq_dim": 32,
"precise_io_type": "async_io",
"precise_file_path": "./precise_codes"
}}'
search_params: '{"ivf": {
"scan_buckets_count": 10
}}'
index_path: "/tmp/sift-128-euclidean/index/hgraph_index"
topk: 10
search_mode: "knn"
range: 0.5
delete_index_after_search: false
1. Describe the bug
When running the
eval_performancescript, the reportedmemory_usagemetric shows an impossibly large value:16777216.00 TB. This is clearly a bug in how memory is being calculated or displayed.2. To Reproduce
eval_performancescript with following configuration.