All-mpnet-base-v2 model fine-tuned for questions clustering

This is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.

This model is named all-mpnet-base-questions-clustering-en since it is a Sentence Transformers model specifically fine-tuned for a questions clustering task. Three public dataset (Quora, WikiAnswer and StackExchange) has been used to enhance the model performance specifically in mapping questions with similar meanings.

Usage (Sentence-Transformers)

Using this model becomes easy when you have sentence-transformers installed:

pip install -U sentence-transformers

Then you can use the model like this:

from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]

model = SentenceTransformer('aiknowyou/all-mpnet-base-questions-clustering-en')
embeddings = model.encode(sentences)
print(embeddings)

Evaluation Results

The present model has been evaluated by employing a test set belonging to the WikiAnswer dataset. The evaluation results are the following:

[ { "epoch": 1, "cossim_accuracy": 0.9931843415744172, "cossim_accuracy_threshold": 0.35143423080444336, "cossim_f1": 0.9897547191636324, "cossim_precision": 0.9913437348280885, "cossim_recall": 0.9881707893839572, "cossim_f1_threshold": 0.35143423080444336, "cossim_ap": 0.9989950013637923, "manhattan_accuracy": 0.9934042015236294, "manhattan_accuracy_threshold": 24.160316467285156, "manhattan_f1": 0.9900818249442103, "manhattan_precision": 0.9920113508380628, "manhattan_recall": 0.9881597905828264, "manhattan_f1_threshold": 24.160316467285156, "manhattan_ap": 0.9990576126715013, "euclidean_accuracy": 0.9931843415744172, "euclidean_accuracy_threshold": 1.1389167308807373, "euclidean_f1": 0.9897547191636324, "euclidean_precision": 0.9913437348280885, "euclidean_recall": 0.9881707893839572, "euclidean_f1_threshold": 1.1389167308807373, "euclidean_ap": 0.9989921332302106, "dot_accuracy": 0.9931843415744172, "dot_accuracy_threshold": 0.35143429040908813, "dot_f1": 0.9897547191636324, "dot_precision": 0.9913437348280885, "dot_recall": 0.9881707893839572, "dot_f1_threshold": 0.35143429040908813, "dot_ap": 0.9989933009226604 } ]

For an automated evaluation of this model, see the Sentence Embeddings Benchmark: https://seb.sbert.net

Training

The model was trained with the parameters:

DataLoader:

torch.utils.data.dataloader.DataLoader of length 34123 with parameters:

{'batch_size': 32, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

Loss:

sentence_transformers.losses.MultipleNegativesRankingLoss.MultipleNegativesRankingLoss with parameters:

{'scale': 20.0, 'similarity_fct': 'cos_sim'}

DataLoader:

torch.utils.data.dataloader.DataLoader of length 51184 with parameters:

{'batch_size': 32, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

Loss:

sentence_transformers.losses.OnlineContrastiveLoss.OnlineContrastiveLoss

Parameters of the fit()-Method:

{
    "epochs": 2,
    "evaluation_steps": 0,
    "evaluator": "sentence_transformers.evaluation.SequentialEvaluator.SequentialEvaluator",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
    "optimizer_params": {
        "lr": 2e-05
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": null,
    "warmup_steps": 1000,
    "weight_decay": 0.01
}

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 384, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
  (2): Normalize()
)

Contribution

Thanks to @tradicio for adding this model.

License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.