Model Card for vectorizer-v1-S-multilingual

This model is a vectorizer developed by Sinequa. It produces an embedding vector given a passage or a query. The passage vectors are stored in our vector index and the query vector is used at query time to look up relevant passages in the index.

Model name: vectorizer-v1-S-multilingual

Supported Languages

The model was trained and tested in the following languages:

• English
• French
• German
• Spanish

Scores

Metric	Value
Relevance (Recall@100)	0.448

Note that the relevance score is computed as an average over 14 retrieval datasets (see details below).

Inference Times

GPU	Quantization type	Batch size 1	Batch size 32
NVIDIA A10	FP16	1 ms	5 ms
NVIDIA A10	FP32	3 ms	14 ms
NVIDIA T4	FP16	1 ms	12 ms
NVIDIA T4	FP32	2 ms	52 ms
NVIDIA L4	FP16	1 ms	5 ms
NVIDIA L4	FP32	2 ms	18 ms

Gpu Memory usage

Quantization type	Memory
FP16	300 MiB
FP32	600 MiB

Note that GPU memory usage only includes how much GPU memory the actual model consumes on an NVIDIA T4 GPU with a batch size of 32. It does not include the fix amount of memory that is consumed by the ONNX Runtime upon initialization which can be around 0.5 to 1 GiB depending on the used GPU.

Requirements

• Minimal Sinequa version: 11.10.0
• Minimal Sinequa version for using FP16 models and GPUs with CUDA compute capability of 8.9+ (like NVIDIA L4): 11.11.0
• Cuda compute capability: above 5.0 (above 6.0 for FP16 use)

Model Details

Overview

• Number of parameters: 39 million
• Base language model: Homegrown Sinequa BERT-Small (Paper) pretrained in the four supported languages
• Insensitive to casing and accents
• Training procedure: Query-passage pairs using in-batch negatives

Training Data

• Natural Questions (Paper, Official Page)
  • Original English dataset
  • Translated datasets for the other three supported languages

Evaluation Metrics

To determine the relevance score, we averaged the results that we obtained when evaluating on the datasets of the BEIR benchmark. Note that all these datasets are in English.

Dataset	Recall@100
Average	0.448
Arguana	0.835
CLIMATE-FEVER	0.350
DBPedia Entity	0.287
FEVER	0.645
FiQA-2018	0.305
HotpotQA	0.396
MS MARCO	0.533
NFCorpus	0.162
NQ	0.701
Quora	0.947
SCIDOCS	0.194
SciFact	0.580
TREC-COVID	0.051
Webis-Touche-2020	0.289

We evaluated the model on the datasets of the MIRACL benchmark to test its multilingual capacities. Note that not all training languages are part of the benchmark, so we only report the metrics for the existing languages.

Language	Recall@100
French	0.583
German	0.524
Spanish	0.483