|
| 1 | +--- |
| 2 | +layout: model |
| 3 | +title: UAE-Large-V1 for Sentence Embeddings |
| 4 | +author: John Snow Labs |
| 5 | +name: uae_large_v1 |
| 6 | +date: 2024-04-05 |
| 7 | +tags: [uae, en, sentence, embeddings, open_source, onnx] |
| 8 | +task: Embeddings |
| 9 | +language: en |
| 10 | +edition: Spark NLP 5.3.3 |
| 11 | +spark_version: 3.0 |
| 12 | +supported: true |
| 13 | +engine: onnx |
| 14 | +annotator: UAEEmbeddings |
| 15 | +article_header: |
| 16 | + type: cover |
| 17 | +use_language_switcher: "Python-Scala-Java" |
| 18 | +--- |
| 19 | + |
| 20 | +## Description |
| 21 | + |
| 22 | +UAE is a novel angle-optimized text embedding model, designed to improve semantic textual |
| 23 | +similarity tasks, which are crucial for Large Language Model (LLM) applications. By |
| 24 | +introducing angle optimization in a complex space, AnglE effectively mitigates saturation of |
| 25 | +the cosine similarity function. |
| 26 | + |
| 27 | +This model is based on UAE-Large-V1 and was orignally exported from https://huggingface.co/WhereIsAI/UAE-Large-V1. Several embedding pooling strategies can be set. Please refer to the class for more information. |
| 28 | + |
| 29 | +## Predicted Entities |
| 30 | + |
| 31 | + |
| 32 | + |
| 33 | +{:.btn-box} |
| 34 | +<button class="button button-orange" disabled>Live Demo</button> |
| 35 | +<button class="button button-orange" disabled>Open in Colab</button> |
| 36 | +[Download](https://s3.amazonaws.com/auxdata.johnsnowlabs.com/public/models/uae_large_v1_en_5.3.3_3.0_1712335736995.zip){:.button.button-orange.button-orange-trans.arr.button-icon} |
| 37 | +[Copy S3 URI](s3://auxdata.johnsnowlabs.com/public/models/uae_large_v1_en_5.3.3_3.0_1712335736995.zip){:.button.button-orange.button-orange-trans.button-icon.button-copy-s3} |
| 38 | + |
| 39 | +## How to use |
| 40 | + |
| 41 | + |
| 42 | + |
| 43 | +<div class="tabs-box" markdown="1"> |
| 44 | +{% include programmingLanguageSelectScalaPythonNLU.html %} |
| 45 | +```python |
| 46 | +import sparknlp |
| 47 | +from sparknlp.base import * |
| 48 | +from sparknlp.annotator import * |
| 49 | +from pyspark.ml import Pipeline |
| 50 | +documentAssembler = DocumentAssembler() \ |
| 51 | + .setInputCol("text") \ |
| 52 | + .setOutputCol("document") |
| 53 | +embeddings = UAEEmbeddings.pretrained() \ |
| 54 | + .setInputCols(["document"]) \ |
| 55 | + .setOutputCol("embeddings") |
| 56 | +embeddingsFinisher = EmbeddingsFinisher() \ |
| 57 | + .setInputCols("embeddings") \ |
| 58 | + .setOutputCols("finished_embeddings") \ |
| 59 | + .setOutputAsVector(True) |
| 60 | +pipeline = Pipeline().setStages([ |
| 61 | + documentAssembler, |
| 62 | + embeddings, |
| 63 | + embeddingsFinisher |
| 64 | +]) |
| 65 | +data = spark.createDataFrame([["hello world", "hello moon"]]).toDF("text") |
| 66 | +result = pipeline.fit(data).transform(data) |
| 67 | +result.selectExpr("explode(finished_embeddings) as result").show(5, 80) |
| 68 | +``` |
| 69 | +```scala |
| 70 | +import spark.implicits._ |
| 71 | +import com.johnsnowlabs.nlp.base.DocumentAssembler |
| 72 | +import com.johnsnowlabs.nlp.annotators.Tokenizer |
| 73 | +import com.johnsnowlabs.nlp.embeddings.UAEEmbeddings |
| 74 | +import com.johnsnowlabs.nlp.EmbeddingsFinisher |
| 75 | +import org.apache.spark.ml.Pipeline |
| 76 | +val documentAssembler = new DocumentAssembler() |
| 77 | + .setInputCol("text") |
| 78 | + .setOutputCol("document") |
| 79 | +val embeddings = UAEEmbeddings.pretrained() |
| 80 | + .setInputCols("document") |
| 81 | + .setOutputCol("UAE_embeddings") |
| 82 | +val embeddingsFinisher = new EmbeddingsFinisher() |
| 83 | + .setInputCols("UAE_embeddings") |
| 84 | + .setOutputCols("finished_embeddings") |
| 85 | + .setOutputAsVector(true) |
| 86 | +val pipeline = new Pipeline().setStages(Array( |
| 87 | + documentAssembler, |
| 88 | + embeddings, |
| 89 | + embeddingsFinisher |
| 90 | +)) |
| 91 | +val data = Seq("hello world", "hello moon").toDF("text") |
| 92 | +val result = pipeline.fit(data).transform(data) |
| 93 | +result.selectExpr("explode(finished_embeddings) as result").show(5, 80) |
| 94 | +``` |
| 95 | +</div> |
| 96 | + |
| 97 | +## Results |
| 98 | + |
| 99 | +```bash |
| 100 | ++--------------------------------------------------------------------------------+ |
| 101 | +| result| |
| 102 | ++--------------------------------------------------------------------------------+ |
| 103 | +|[0.50387806, 0.5861606, 0.35129607, -0.76046336, -0.32446072, -0.117674336, 0...| |
| 104 | +|[0.6660665, 0.961762, 0.24854276, -0.1018044, -0.6569202, 0.027635604, 0.1915...| |
| 105 | ++--------------------------------------------------------------------------------+ |
| 106 | +``` |
| 107 | +
|
| 108 | +{:.model-param} |
| 109 | +## Model Information |
| 110 | +
|
| 111 | +{:.table-model} |
| 112 | +|---|---| |
| 113 | +|Model Name:|uae_large_v1| |
| 114 | +|Compatibility:|Spark NLP 5.3.3+| |
| 115 | +|License:|Open Source| |
| 116 | +|Edition:|Official| |
| 117 | +|Input Labels:|[document]| |
| 118 | +|Output Labels:|[embeddings]| |
| 119 | +|Language:|en| |
| 120 | +|Size:|1.2 GB| |
0 commit comments