|
|
<!DOCTYPE html> |
|
|
<html lang="en"> |
|
|
<head> |
|
|
<meta charset="utf-8"> |
|
|
<meta name="viewport" content="width=device-width, initial-scale=1.0"> |
|
|
<title>SBERT.net Models</title> |
|
|
|
|
|
<script src="https://cdnjs.cloudflare.com/ajax/libs/vue/2.6.12/vue.min.js" integrity="sha512-BKbSR+cfyxLdMAsE0naLReFSLg8/pjbgfxHh/k/kUC82Hy7r6HtR5hLhobaln2gcTvzkyyehrdREdjpsQwy2Jw==" crossorigin="anonymous" referrerpolicy="no-referrer"></script> |
|
|
|
|
|
|
|
|
|
|
|
<link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.4.1/css/bootstrap.min.css" |
|
|
integrity="sha384-Vkoo8x4CGsO3+Hhxv8T/Q5PaXtkKtu6ug5TOeNV6gBiFeWPGFN9MuhOf23Q9Ifjh" crossorigin="anonymous"> |
|
|
|
|
|
<script src="https://code.jquery.com/jquery-3.4.1.slim.min.js" |
|
|
integrity="sha384-J6qa4849blE2+poT4WnyKhv5vZF5SrPo0iEjwBvKU7imGFAV0wwj1yYfoRSJoZ+n" |
|
|
crossorigin="anonymous"></script> |
|
|
<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/umd/popper.min.js" |
|
|
integrity="sha384-Q6E9RHvbIyZFJoft+2mJbHaEWldlvI9IOYy5n3zV9zzTtmI3UksdQRVvoxMfooAo" |
|
|
crossorigin="anonymous"></script> |
|
|
<script src="https://stackpath.bootstrapcdn.com/bootstrap/4.4.1/js/bootstrap.min.js" |
|
|
integrity="sha384-wfSDF2E50Y2D1uUdj0O3uMBJnjuUD4Ih7YwaYd1iqfktj0Uod8GCExl3Og8ifwB6" |
|
|
crossorigin="anonymous"></script> |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.15.3/css/all.min.css" |
|
|
integrity="sha512-iBBXm8fW90+nuLcSKlbmrPcLa0OT92xO1BIsZ+ywDWZCvqsWgccV3gFoRBv0z+8dLJgyAHIhR35VZc2oM/gI1w==" |
|
|
crossorigin="anonymous" referrerpolicy="no-referrer"/> |
|
|
|
|
|
|
|
|
<script src="https://cdnjs.cloudflare.com/ajax/libs/lodash.js/4.17.21/lodash.min.js" |
|
|
integrity="sha512-WFN04846sdKMIP5LKNphMaWzU7YpMyCU245etK3g/2ARYbPK9Ub18eG+ljU96qKRCWh+quCY7yefSmlkQw1ANQ==" |
|
|
crossorigin="anonymous" referrerpolicy="no-referrer"></script> |
|
|
|
|
|
<style> |
|
|
.fa-active { |
|
|
color: #337ab7; |
|
|
} |
|
|
|
|
|
|
|
|
.header-cell { |
|
|
cursor: pointer; |
|
|
} |
|
|
|
|
|
.models-table thead th { |
|
|
position: sticky; |
|
|
top: 0; |
|
|
z-index: 1; |
|
|
background-color: #ffffff; |
|
|
} |
|
|
|
|
|
|
|
|
.info-icon { |
|
|
color: #cccccc; |
|
|
} |
|
|
|
|
|
.info-icon-disabled { |
|
|
color: #cccccc; |
|
|
} |
|
|
|
|
|
.link-active, .toggle-active, .info-icon-active, .info-icon:hover { |
|
|
color: #0056b3; |
|
|
} |
|
|
|
|
|
.info-icon-model { |
|
|
padding-left: 10px; |
|
|
} |
|
|
|
|
|
.bs-popover-auto[x-placement^=bottom], .bs-popover-bottom { |
|
|
margin-top: .5rem; |
|
|
} |
|
|
|
|
|
.popover { |
|
|
max-width: 400px; |
|
|
} |
|
|
|
|
|
.toggle-link { |
|
|
text-decoration: none; |
|
|
cursor: pointer; |
|
|
padding: 10px; |
|
|
} |
|
|
|
|
|
.toggle-link:hover, .toggle-link:focus { |
|
|
text-decoration: none; |
|
|
} |
|
|
|
|
|
.toggle-inactive, .toggle-inactive:hover, .toggle-inactive:focus { |
|
|
color: #333; |
|
|
} |
|
|
|
|
|
</style> |
|
|
</head> |
|
|
<body> |
|
|
|
|
|
<div id="app"> |
|
|
<div class="text-right p-2"> |
|
|
<span class="toggle-link" v-bind:class="{'toggle-active': show_all_models, 'toggle-inactive': !show_all_models}" @click="show_all_models = !show_all_models"> |
|
|
<span v-if="show_all_models">All models</span><span v-else>All models</span> |
|
|
<i class="fas" v-bind:class="{'toggle-active': show_all_models, 'fa-toggle-on': show_all_models, 'fa-toggle-off': !show_all_models}"></i> |
|
|
</span> |
|
|
</div> |
|
|
|
|
|
<table class="table table-sm"> |
|
|
<thead> |
|
|
<tr> |
|
|
<th class="header-cell" @click="sortAsc = (sortBy=='name') ? sortAsc = !sortAsc : false; sortBy='name'"> |
|
|
<i class="fas fa-active" v-if="sortBy == 'name'" v-bind:class="{ 'fa-sort-amount-up': !sortAsc, 'fa-sort-amount-down-alt': sortAsc }"></i> |
|
|
Model Name |
|
|
</th> |
|
|
<th class="header-cell text-center" @click="sortAsc = (sortBy=='sentence_performance') ? sortAsc = !sortAsc : false; sortBy='sentence_performance'"> |
|
|
<i class="fas fa-active" v-if="sortBy == 'sentence_performance'" v-bind:class="{ 'fa-sort-amount-up': !sortAsc, 'fa-sort-amount-down-alt': sortAsc }"></i> |
|
|
Performance Sentence Embeddings (14 Datasets) |
|
|
<span class="info-icon" data-trigger="hover" data-toggle="popover" title="Performance Sentence Embeddings" data-content="Average performance on encoding sentences over 14 diverse tasks from different domains.<br>Higher = Better" data-html="true" data-placement="bottom"><i class="fas fa-info-circle"></i></span> |
|
|
</th> |
|
|
<th class="header-cell text-center" @click="sortAsc = (sortBy=='semantic_search') ? sortAsc = !sortAsc : false; sortBy='semantic_search'"> |
|
|
<i class="fas fa-active" v-if="sortBy == 'semantic_search'" v-bind:class="{ 'fa-sort-amount-up': !sortAsc, 'fa-sort-amount-down-alt': sortAsc }"></i> |
|
|
Performance Semantic Search (6 Datasets) |
|
|
<span class="info-icon" data-trigger="hover" data-toggle="popover" title="Performance Semantic Search" data-content="Performance on 6 diverse tasks for semantic search: Encoding of queries / questions and paragraphs up to 512 word pieces.<br>Higher = Better." data-html="true" data-placement="bottom"><i class="fas fa-info-circle"></i></span> |
|
|
</th> |
|
|
<th class="header-cell text-center" @click="sortAsc = (sortBy=='avg_performance') ? sortAsc = !sortAsc : false; sortBy='avg_performance'"> |
|
|
<i class="fas fa-active" v-if="sortBy == 'avg_performance'" v-bind:class="{ 'fa-sort-amount-up': !sortAsc, 'fa-sort-amount-down-alt': sortAsc }"></i> |
|
|
Avg. Performance |
|
|
<span class="info-icon" data-trigger="hover" data-toggle="popover" title="Average Performance" data-content="Average of sentence performance and semantic search performance.<br>Higher = Better." data-html="true" data-placement="bottom"><i class="fas fa-info-circle"></i></span> |
|
|
</th> |
|
|
<th class="header-cell text-center" @click="sortAsc = (sortBy=='speed') ? sortAsc = !sortAsc : false; sortBy='speed'"> |
|
|
<i class="fas fa-active" v-if="sortBy == 'speed'" v-bind:class="{ 'fa-sort-amount-up': !sortAsc, 'fa-sort-amount-down-alt': sortAsc }"></i> |
|
|
Speed |
|
|
<span class="info-icon" data-trigger="hover" data-toggle="popover" title="Encoding Speed" data-content="Encoding speed (sentences / sec) on a V100 GPU.<br>Higher = Better" data-html="true" data-placement="bottom"><i class="fas fa-info-circle"></i></span> |
|
|
</th> |
|
|
<th class="header-cell text-center" @click="sortAsc = (sortBy=='size') ? sortAsc = !sortAsc : false; sortBy='size'"> |
|
|
<i class="fas fa-active" v-if="sortBy == 'size'" v-bind:class="{ 'fa-sort-amount-up': !sortAsc, 'fa-sort-amount-down-alt': sortAsc }"></i> |
|
|
Model Size |
|
|
<span class="info-icon" data-trigger="hover" data-toggle="popover" title="Size" data-content="Size (in MB) of the model." data-html="true" data-placement="bottom"><i class="fas fa-info-circle"></i></span> |
|
|
</th> |
|
|
</tr> |
|
|
</thead> |
|
|
<tbody> |
|
|
<template v-for="item in sortedModels"> |
|
|
<tr v-on:click="item.show_details = !item.show_details" style="cursor: pointer"> |
|
|
<td style="white-space: nowrap;"> |
|
|
{{ item.name }} |
|
|
<span class="info-icon-model" v-bind:class="{'info-icon-active': item.show_details, 'info-icon-disabled': !item.show_details}" ><i class="fas fa-info-circle"></i></span> |
|
|
</td> |
|
|
<td class="text-center">{{ item.sentence_performance > 0 ? item.sentence_performance.toFixed(2) : "" }}</td> |
|
|
<td class="text-center">{{ item.semantic_search > 0 ? item.semantic_search.toFixed(2) : "" }}</td> |
|
|
<td class="text-center">{{ (item.sentence_performance > 0 && item.semantic_search > 0) ? item.avg_performance.toFixed(2) : "" }}</td> |
|
|
<td class="text-center">{{ item.speed }}</td> |
|
|
<td class="text-center">{{ item.size }} MB</td> |
|
|
</tr> |
|
|
<tr v-if="item.show_details"> |
|
|
<td colspan="6" style="padding-left: 20px"> |
|
|
<table class="table table-sm" style="width: 100%; font-size: 0.9em;"> |
|
|
<thead> |
|
|
<tr> |
|
|
<td colspan="2"> |
|
|
<b>{{ item.name }}</b> |
|
|
<button title="Copy model name" type="button" class="btn btn-link p-0" v-on:click="copyClipboard(item.name)" data-toggle="tooltip" data-placement="bottom" data-trigger="hover" :id="item.name+'-copy-btn'" style="border: 0;"> |
|
|
<svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> |
|
|
</button> |
|
|
|
|
|
<br> |
|
|
|
|
|
|
|
|
</td> |
|
|
</tr> |
|
|
</thead> |
|
|
<tbody> |
|
|
<tr v-if="item.hasOwnProperty('description')"> |
|
|
<th>Description:</th> |
|
|
<td>{{item.description}}</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Base Model:</th> |
|
|
<td><span v-html="item.base_model"></span></td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Max Sequence Length:</th> |
|
|
<td>{{item.max_seq_length || ''}}</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Dimensions:</th> |
|
|
<td>{{item.dim }}</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th style="width: 220px;">Normalized Embeddings:</th> |
|
|
<td>{{item.normalized_embeddings}}</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Suitable Score Functions:</th> |
|
|
<td> |
|
|
<span v-html="getScoreFunction(item.score_functions)"></span> |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Size:</th> |
|
|
<td>{{item.size}} MB</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Pooling:</th> |
|
|
<td>{{item.pooling}}</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Training Data:</th> |
|
|
<td>{{item.training_data}}</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<th>Model Card:</th> |
|
|
<td><a :href="'https://huggingface.co/sentence-transformers/'+item.name" target="_blank">https://huggingface.co/sentence-transformers/{{item.name}}</a></td> |
|
|
</tr> |
|
|
</tbody> |
|
|
</table> |
|
|
|
|
|
</td> |
|
|
</tr> |
|
|
</template> |
|
|
|
|
|
</tbody> |
|
|
</table> |
|
|
</div> |
|
|
|
|
|
<script> |
|
|
|
|
|
var app = new Vue({ |
|
|
el: '#app', |
|
|
data: { |
|
|
show_all_models: false, |
|
|
models: [ |
|
|
{ |
|
|
"name": "average_word_embeddings_glove.6B.300d", |
|
|
"base_model": "Word Embeddings: GloVe", |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "-", |
|
|
"sentence_performance": 49.79, |
|
|
"semantic_search": 22.71, |
|
|
"speed": 34000, |
|
|
"size": 420, |
|
|
"dim": 300, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"] |
|
|
}, |
|
|
{ |
|
|
"name": "average_word_embeddings_komninos", |
|
|
"base_model": "Word Embeddings: Komninos et al.", |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "-", |
|
|
"sentence_performance": 51.13, |
|
|
"semantic_search": 21.64, |
|
|
"speed": 22000, |
|
|
"size": 240, |
|
|
"dim": 300, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"] |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-MiniLM-L3-v2", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/MiniLM-L6-H384-uncased" target="_blank">nreimers/MiniLM-L3-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "AllNLI, sentence-compression, SimpleWiki, altlex, msmarco-triplets, quora_duplicates, coco_captions,flickr30k_captions, yahoo_answers_title_question, S2ORC_citation_pairs, stackexchange_duplicate_questions, wiki-atomic-edits", |
|
|
"sentence_performance": 62.29, |
|
|
"semantic_search": 39.19, |
|
|
"speed": 19000, |
|
|
"size": 61, |
|
|
"dim": 384, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-MiniLM-L6-v2", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/MiniLM-L6-H384-uncased" target="_blank">nreimers/MiniLM-L6-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "AllNLI, sentence-compression, SimpleWiki, altlex, msmarco-triplets, quora_duplicates, coco_captions,flickr30k_captions, yahoo_answers_title_question, S2ORC_citation_pairs, stackexchange_duplicate_questions, wiki-atomic-edits", |
|
|
"sentence_performance": 64.82, |
|
|
"semantic_search": 40.31, |
|
|
"speed": 14200, |
|
|
"size": 80, |
|
|
"dim": 384, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"] |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-MiniLM-L12-v2", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/MiniLM-L12-H384-uncased" target="_blank">microsoft/MiniLM-L12-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "AllNLI, sentence-compression, SimpleWiki, altlex, msmarco-triplets, quora_duplicates, coco_captions,flickr30k_captions, yahoo_answers_title_question, S2ORC_citation_pairs, stackexchange_duplicate_questions, wiki-atomic-edits", |
|
|
"sentence_performance": 66.01, |
|
|
"semantic_search": 43.01, |
|
|
"speed": 7500, |
|
|
"size": 120, |
|
|
"dim": 384, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"] |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-distilroberta-base-v2", |
|
|
"base_model": '<a href="https://huggingface.co/distilroberta-base" target="_blank">distilroberta-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "AllNLI, sentence-compression, SimpleWiki, altlex, msmarco-triplets, quora_duplicates, coco_captions,flickr30k_captions, yahoo_answers_title_question, S2ORC_citation_pairs, stackexchange_duplicate_questions, wiki-atomic-edits", |
|
|
"sentence_performance": 66.27, |
|
|
"semantic_search": 43.10, |
|
|
"speed": 4000, |
|
|
"size": 290, |
|
|
"dim": 768, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"] |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-TinyBERT-L6-v2", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/TinyBERT_L-6_H-768_v2" target="_blank">nreimers/TinyBERT_L-6_H-768_v2</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "AllNLI, sentence-compression, SimpleWiki, altlex, msmarco-triplets, quora_duplicates, coco_captions,flickr30k_captions, yahoo_answers_title_question, S2ORC_citation_pairs, stackexchange_duplicate_questions, wiki-atomic-edits", |
|
|
"sentence_performance": 66.19, |
|
|
"semantic_search": 41.07, |
|
|
"speed": 4500, |
|
|
"size": 240, |
|
|
"dim": 768, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"] |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-mpnet-base-v2", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/mpnet-base" target="_blank">microsoft/mpnet-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "AllNLI, sentence-compression, SimpleWiki, altlex, msmarco-triplets, quora_duplicates, coco_captions,flickr30k_captions, yahoo_answers_title_question, S2ORC_citation_pairs, stackexchange_duplicate_questions, wiki-atomic-edits", |
|
|
"sentence_performance": 67.97, |
|
|
"semantic_search": 47.43, |
|
|
"speed": 2800, |
|
|
"size": 420, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"] |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-albert-small-v2", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/albert-small-v2" target="_blank">nreimers/albert-small-v2</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "AllNLI, sentence-compression, SimpleWiki, altlex, msmarco-triplets, quora_duplicates, coco_captions,flickr30k_captions, yahoo_answers_title_question, S2ORC_citation_pairs, stackexchange_duplicate_questions, wiki-atomic-edits", |
|
|
"sentence_performance": 64.46, |
|
|
"semantic_search": 40.04, |
|
|
"speed": 5000, |
|
|
"size": 43, |
|
|
"dim": 768, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-multilingual-mpnet-base-v2", |
|
|
"base_model": "Teacher: paraphrase-mpnet-base-v2; Student: xlm-roberta-base", |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "Multi-lingual model of paraphrase-mpnet-base-v2, extended to 50+ languages.", |
|
|
"sentence_performance": 65.83, |
|
|
"semantic_search": 41.68, |
|
|
"speed": 2500, |
|
|
"size": 970, |
|
|
"dim": 768, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "paraphrase-multilingual-MiniLM-L12-v2", |
|
|
"base_model": "Teacher: paraphrase-MiniLM-L12-v2; Student: microsoft/Multilingual-MiniLM-L12-H384", |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "Multi-lingual model of paraphrase-multilingual-MiniLM-L12-v2, extended to 50+ languages.", |
|
|
"sentence_performance": 64.25, |
|
|
"semantic_search": 39.19, |
|
|
"speed": 7500, |
|
|
"size": 420, |
|
|
"dim": 384, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "distiluse-base-multilingual-cased-v1", |
|
|
"base_model": "Teacher: mUSE; Student: distilbert-base-multilingual", |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "Multi-Lingual model of Universal Sentence Encoder for 15 languages: Arabic, Chinese, Dutch, English, French, German, Italian, Korean, Polish, Portuguese, Russian, Spanish, Turkish.", |
|
|
"sentence_performance": 61.30, |
|
|
"semantic_search": 29.87, |
|
|
"speed": 4000, |
|
|
"size": 480, |
|
|
"dim": 512, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "distiluse-base-multilingual-cased-v2", |
|
|
"base_model": "Teacher: mUSE; Student: distilbert-base-multilingual", |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "Multi-Lingual model of Universal Sentence Encoder for 50 languages.", |
|
|
"sentence_performance": 60.18, |
|
|
"semantic_search": 27.35, |
|
|
"speed": 4000, |
|
|
"size": 480, |
|
|
"dim": 512, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["cos"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "all-distilroberta-v1", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/distilroberta-base" target="_blank">distilroberta-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 68.73, |
|
|
"semantic_search": 50.94, |
|
|
"speed": 4000, |
|
|
"size": 290, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "all-MiniLM-L6-v1", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/MiniLM-L6-H384-uncased" target="_blank">nreimers/MiniLM-L6-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 68.03, |
|
|
"semantic_search": 48.07, |
|
|
"speed": 14200, |
|
|
"size": 80, |
|
|
"dim": 384, |
|
|
"max_seq_length": 128, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"] |
|
|
}, |
|
|
{ |
|
|
"name": "all-MiniLM-L6-v2", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/MiniLM-L6-H384-uncased" target="_blank">nreimers/MiniLM-L6-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 68.06, |
|
|
"semantic_search": 49.54, |
|
|
"speed": 14200, |
|
|
"size": 80, |
|
|
"dim": 384, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "all-MiniLM-L12-v1", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/MiniLM-L12-H384-uncased" target="_blank">microsoft/MiniLM-L12-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 68.83, |
|
|
"semantic_search": 50.78, |
|
|
"speed": 7500, |
|
|
"size": 120, |
|
|
"dim": 384, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"] |
|
|
}, |
|
|
{ |
|
|
"name": "all-MiniLM-L12-v2", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/MiniLM-L12-H384-uncased" target="_blank">microsoft/MiniLM-L12-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 68.7, |
|
|
"semantic_search": 50.82, |
|
|
"speed": 7500, |
|
|
"size": 120, |
|
|
"dim": 384, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "all-mpnet-base-v1", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/mpnet-base" target="_blank">microsoft/mpnet-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 69.98, |
|
|
"semantic_search": 54.69, |
|
|
"speed": 2800, |
|
|
"size": 420, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"] |
|
|
}, |
|
|
{ |
|
|
"name": "all-mpnet-base-v2", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/mpnet-base" target="_blank">microsoft/mpnet-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 69.57, |
|
|
"semantic_search": 57.02, |
|
|
"speed": 2800, |
|
|
"size": 420, |
|
|
"dim": 768, |
|
|
"max_seq_length": 384, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "all-roberta-large-v1", |
|
|
"description": "All-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/roberta-large" target="_blank">roberta-large</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "1B+ training pairs. For details, see model card.", |
|
|
"sentence_performance": 70.23, |
|
|
"semantic_search": 53.05, |
|
|
"speed": 800, |
|
|
"size": 1360, |
|
|
"dim": 1024, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"] |
|
|
}, |
|
|
{ |
|
|
"name": "multi-qa-MiniLM-L6-dot-v1", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on a large and diverse set of (question, answer) pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/MiniLM-L6-H384-uncased" target="_blank">nreimers/MiniLM-L6-H384-uncased</a>', |
|
|
"pooling": "CLS Pooling", |
|
|
"training_data": "215M (question, answer) pairs from diverse sources.", |
|
|
"sentence_performance": 63.90, |
|
|
"semantic_search": 49.19, |
|
|
"speed": 14200, |
|
|
"size": 80, |
|
|
"dim": 384, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["dot"] |
|
|
}, |
|
|
{ |
|
|
"name": "multi-qa-MiniLM-L6-cos-v1", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on a large and diverse set of (question, answer) pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/nreimers/MiniLM-L6-H384-uncased" target="_blank">nreimers/MiniLM-L6-H384-uncased</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "215M (question, answer) pairs from diverse sources.", |
|
|
"sentence_performance": 64.33, |
|
|
"semantic_search": 51.83, |
|
|
"speed": 14200, |
|
|
"size": 80, |
|
|
"dim": 384, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "multi-qa-distilbert-dot-v1", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on a large and diverse set of (question, answer) pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/distilbert-base" target="_blank">distilbert-base</a>', |
|
|
"pooling": "CLS Pooling", |
|
|
"training_data": "215M (question, answer) pairs from diverse sources.", |
|
|
"sentence_performance": 66.67, |
|
|
"semantic_search": 52.51, |
|
|
"speed": 4000, |
|
|
"size": 250, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["dot"] |
|
|
}, |
|
|
{ |
|
|
"name": "multi-qa-distilbert-cos-v1", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on a large and diverse set of (question, answer) pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/distilbert-base" target="_blank">distilbert-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "215M (question, answer) pairs from diverse sources.", |
|
|
"sentence_performance": 65.98, |
|
|
"semantic_search": 52.83, |
|
|
"speed": 4000, |
|
|
"size": 250, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "multi-qa-mpnet-base-dot-v1", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on a large and diverse set of (question, answer) pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/mpnet-base" target="_blank"microsoft/mpnet-base</a>', |
|
|
"pooling": "CLS Pooling", |
|
|
"training_data": "215M (question, answer) pairs from diverse sources.", |
|
|
"sentence_performance": 66.76, |
|
|
"semantic_search": 57.60, |
|
|
"speed": 2800, |
|
|
"size": 420, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["dot"], |
|
|
"recommended_model": true |
|
|
}, |
|
|
{ |
|
|
"name": "multi-qa-mpnet-base-cos-v1", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on a large and diverse set of (question, answer) pairs.", |
|
|
"base_model": '<a href="https://huggingface.co/microsoft/mpnet-base" target="_blank">microsoft/mpnet-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "215M (question, answer) pairs from diverse sources.", |
|
|
"sentence_performance": 66.29, |
|
|
"semantic_search": 57.46, |
|
|
"speed": 2800, |
|
|
"size": 420, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "msmarco-distilbert-dot-v5", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on the MS MARCO passages dataset.", |
|
|
"base_model": '<a href="https://huggingface.co/distilbert-base" target="_blank">distilbert-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "500k (query, answer) pairs from MS MARCO Passages dataset.", |
|
|
"sentence_performance": 61.84, |
|
|
"semantic_search": 49.47, |
|
|
"speed": 4000, |
|
|
"size": 250, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["dot"], |
|
|
}, |
|
|
{ |
|
|
"name": "msmarco-bert-base-dot-v5", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on the MS MARCO passages dataset.", |
|
|
"base_model": '<a href="https://huggingface.co/Luyu/co-condenser-marco" target="_blank">Luyu/co-condenser-marco</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "500k (query, answer) pairs from MS MARCO Passages dataset.", |
|
|
"sentence_performance": 62.68, |
|
|
"semantic_search": 52.11, |
|
|
"speed": 2800, |
|
|
"size": 420, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["dot"], |
|
|
}, |
|
|
{ |
|
|
"name": "msmarco-distilbert-base-tas-b", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages. It was trained on the MS MARCO passages dataset.", |
|
|
"base_model": '<a href="https://huggingface.co/distilbert-base" target="_blank">distilbert-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "500k (query, answer) pairs from MS MARCO Passages dataset.", |
|
|
"sentence_performance": 62.57, |
|
|
"semantic_search": 49.25, |
|
|
"speed": 4000, |
|
|
"size": 250, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": false, |
|
|
"score_functions": ["dot"], |
|
|
}, |
|
|
{ |
|
|
"name": "sentence-t5-base", |
|
|
"description": "This model was tuned for sentence similarity tasks.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-base" target="_blank">t5-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities.", |
|
|
"sentence_performance": 67.84, |
|
|
"semantic_search": 44.63, |
|
|
"speed": 2500, |
|
|
"size": 210, |
|
|
"dim": 768, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "sentence-t5-large", |
|
|
"description": "This model was tuned for sentence similarity tasks.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-large" target="_blank">t5-large</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities.", |
|
|
"sentence_performance": 68.74, |
|
|
"semantic_search": 49.05, |
|
|
"speed": 800, |
|
|
"size": 640, |
|
|
"dim": 768, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "sentence-t5-xl", |
|
|
"description": "This model was tuned for sentence similarity tasks.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-3b" target="_blank">t5-3b</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities.", |
|
|
"sentence_performance": 69.23, |
|
|
"semantic_search": 51.19, |
|
|
"speed": 230, |
|
|
"size": 2370, |
|
|
"dim": 768, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "sentence-t5-xxl", |
|
|
"description": "This model was tuned for sentence similarity tasks.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-11b" target="_blank">t5-11b</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities.", |
|
|
"sentence_performance": 70.88, |
|
|
"semantic_search": 54.40, |
|
|
"speed": 50, |
|
|
"size": 9230, |
|
|
"dim": 768, |
|
|
"max_seq_length": 256, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "gtr-t5-base", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-base" target="_blank">t5-base</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities and then on MS-MARCO.", |
|
|
"sentence_performance": 67.65, |
|
|
"semantic_search": 51.15, |
|
|
"speed": 2500, |
|
|
"size": 210, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "gtr-t5-large", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-large" target="_blank">t5-large</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities and then on MS-MARCO.", |
|
|
"sentence_performance": 69.90, |
|
|
"semantic_search": 54.85, |
|
|
"speed": 800, |
|
|
"size": 640, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "gtr-t5-xl", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-3b" target="_blank">t5-3b</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities and then on MS-MARCO.", |
|
|
"sentence_performance": 69.88, |
|
|
"semantic_search": 55.88, |
|
|
"speed": 230, |
|
|
"size": 2370, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
{ |
|
|
"name": "gtr-t5-xxl", |
|
|
"description": "This model was tuned for semantic search: Given a query/question, if can find relevant passages.", |
|
|
"base_model": '<a href="https://huggingface.co/t5-11b" target="_blank">t5-11b</a>', |
|
|
"pooling": "Mean Pooling", |
|
|
"training_data": "2B question-answer pairs from diverse online communities and then on MS-MARCO.", |
|
|
"sentence_performance": 70.73, |
|
|
"semantic_search": 55.76, |
|
|
"speed": 50, |
|
|
"size": 9230, |
|
|
"dim": 768, |
|
|
"max_seq_length": 512, |
|
|
"show_details": false, |
|
|
"normalized_embeddings": true, |
|
|
"score_functions": ["dot", "cos", "eucl"], |
|
|
}, |
|
|
|
|
|
|
|
|
], |
|
|
sortBy: 'avg_performance', |
|
|
sortAsc: false |
|
|
}, |
|
|
methods: { |
|
|
copyClipboard: function(msg) { |
|
|
console.log(msg); |
|
|
const clipboardData = window.clipboardData || navigator.clipboard; |
|
|
clipboardData.writeText(msg); |
|
|
const copy_btn = $("#"+msg+"-copy-btn") |
|
|
copy_btn.tooltip('show'); |
|
|
copy_btn.tooltip('hide').attr('data-original-title', "Copied").tooltip('show'); |
|
|
|
|
|
setTimeout(function(){ copy_btn.tooltip('hide').attr('data-original-title', "Copy model name"); }, 1000); |
|
|
}, |
|
|
getScoreFunction: function(score_fct) { |
|
|
let output_html = []; |
|
|
for(let fct of score_fct) { |
|
|
switch(fct) { |
|
|
case "dot": |
|
|
output_html.push("dot-product (<code>util.dot_score</code>)"); |
|
|
break; |
|
|
case "cos": |
|
|
output_html.push("cosine-similarity (<code>util.cos_sim</code>)"); |
|
|
break; |
|
|
case "eucl": |
|
|
output_html.push("euclidean distance") |
|
|
break; |
|
|
default: |
|
|
output_html.push(fct) |
|
|
} |
|
|
} |
|
|
return output_html.join(", "); |
|
|
} |
|
|
}, |
|
|
created: function() { |
|
|
let uri = window.location.search.substring(1); |
|
|
let params = new URLSearchParams(uri); |
|
|
if(params.get("model_name") !== null) { |
|
|
this.show_all_models = true; |
|
|
} |
|
|
}, |
|
|
computed: { |
|
|
sortedModels: function() { |
|
|
|
|
|
let models_ext = this.models.map(function(elem, index) { elem.avg_performance = (elem.sentence_performance + elem.semantic_search)/2.0; return elem;} ); |
|
|
|
|
|
if(!this.show_all_models) { |
|
|
models_ext = models_ext.filter(item => item.recommended_model); |
|
|
} |
|
|
|
|
|
return _.orderBy(models_ext, (item) => item[this.sortBy] || (this.sortAsc ? 9999 : -9999), this.sortAsc ? 'asc' : 'desc') |
|
|
} |
|
|
} |
|
|
}) |
|
|
</script> |
|
|
|
|
|
<script> |
|
|
$(function () { |
|
|
$('[data-toggle="popover"]').popover() |
|
|
}); |
|
|
$(function () { |
|
|
$('[data-toggle="tooltip"]').tooltip() |
|
|
}) |
|
|
</script> |
|
|
|
|
|
|
|
|
</body> |
|
|
</html> |
|
|
|
