TorchSharpCatalog.SentenceSimilarity Method
Definition
Important
Some information relates to prerelease product that may be substantially modified before it’s released. Microsoft makes no warranties, express or implied, with respect to the information provided here.
Overloads
| SentenceSimilarity(RegressionCatalog+RegressionTrainers, SentenceSimilarityTrainer+SentenceSimilarityOptions) |
Fine tune a NAS-BERT model for NLP sentence Similarity. The limit for any sentence is 512 tokens. Each word typically will map to a single token, and we automatically add 2 specical tokens (a start token and a separator token) so in general this limit will be 510 words for all sentences. |
| SentenceSimilarity(RegressionCatalog+RegressionTrainers, String, String, String, String, Int32, Int32, BertArchitecture, IDataView) |
Fine tune a NAS-BERT model for NLP sentence Similarity. The limit for any sentence is 512 tokens. Each word typically will map to a single token, and we automatically add 2 specical tokens (a start token and a separator token) so in general this limit will be 510 words for all sentences. |
SentenceSimilarity(RegressionCatalog+RegressionTrainers, SentenceSimilarityTrainer+SentenceSimilarityOptions)
Fine tune a NAS-BERT model for NLP sentence Similarity. The limit for any sentence is 512 tokens. Each word typically will map to a single token, and we automatically add 2 specical tokens (a start token and a separator token) so in general this limit will be 510 words for all sentences.
public static Microsoft.ML.TorchSharp.NasBert.SentenceSimilarityTrainer SentenceSimilarity (this Microsoft.ML.RegressionCatalog.RegressionTrainers catalog, Microsoft.ML.TorchSharp.NasBert.SentenceSimilarityTrainer.SentenceSimilarityOptions options);static member SentenceSimilarity : Microsoft.ML.RegressionCatalog.RegressionTrainers * Microsoft.ML.TorchSharp.NasBert.SentenceSimilarityTrainer.SentenceSimilarityOptions -> Microsoft.ML.TorchSharp.NasBert.SentenceSimilarityTrainer<Extension()>
Public Function SentenceSimilarity (catalog As RegressionCatalog.RegressionTrainers, options As SentenceSimilarityTrainer.SentenceSimilarityOptions) As SentenceSimilarityTrainerParameters
The transform's catalog.
Advanced Options
Returns
Applies to
SentenceSimilarity(RegressionCatalog+RegressionTrainers, String, String, String, String, Int32, Int32, BertArchitecture, IDataView)
Fine tune a NAS-BERT model for NLP sentence Similarity. The limit for any sentence is 512 tokens. Each word typically will map to a single token, and we automatically add 2 specical tokens (a start token and a separator token) so in general this limit will be 510 words for all sentences.
public static Microsoft.ML.TorchSharp.NasBert.SentenceSimilarityTrainer SentenceSimilarity (this Microsoft.ML.RegressionCatalog.RegressionTrainers catalog, string labelColumnName = "Label", string scoreColumnName = "Score", string sentence1ColumnName = "Sentence1", string sentence2ColumnName = "Sentence2", int batchSize = 32, int maxEpochs = 10, Microsoft.ML.TorchSharp.NasBert.BertArchitecture architecture = Microsoft.ML.TorchSharp.NasBert.BertArchitecture.Roberta, Microsoft.ML.IDataView validationSet = default);static member SentenceSimilarity : Microsoft.ML.RegressionCatalog.RegressionTrainers * string * string * string * string * int * int * Microsoft.ML.TorchSharp.NasBert.BertArchitecture * Microsoft.ML.IDataView -> Microsoft.ML.TorchSharp.NasBert.SentenceSimilarityTrainer<Extension()>
Public Function SentenceSimilarity (catalog As RegressionCatalog.RegressionTrainers, Optional labelColumnName As String = "Label", Optional scoreColumnName As String = "Score", Optional sentence1ColumnName As String = "Sentence1", Optional sentence2ColumnName As String = "Sentence2", Optional batchSize As Integer = 32, Optional maxEpochs As Integer = 10, Optional architecture As BertArchitecture = Microsoft.ML.TorchSharp.NasBert.BertArchitecture.Roberta, Optional validationSet As IDataView = Nothing) As SentenceSimilarityTrainerParameters
The transform's catalog.
- labelColumnName
- String
Name of the label column. Column should be a float type.
- scoreColumnName
- String
Name of the score column.
- sentence1ColumnName
- String
Name of the column for the first sentence.
- sentence2ColumnName
- String
Name of the column for the second sentence. Only required if your NLP classification requires sentence pairs.
- batchSize
- Int32
Number of rows in the batch.
- maxEpochs
- Int32
Maximum number of times to loop through your training set.
- architecture
- BertArchitecture
Architecture for the model. Defaults to Roberta.
- validationSet
- IDataView
The validation set used while training to improve model quality.