🚀 All-mpnet-base-v2模型:用於問題聚類的微調模型
本模型基於 sentence-transformers 構建,可將句子和段落映射到 768 維的密集向量空間,適用於聚類或語義搜索等任務。
該模型名為 all-mpnet-base-questions-clustering-en,是專門為問題聚類任務微調的 Sentence Transformers 模型。它使用了三個公開數據集(Quora、WikiAnswer 和 StackExchange)進行訓練,以提升在映射相似問題時的性能。
🚀 快速開始
📦 安裝指南
若已安裝 sentence-transformers,使用本模型將十分便捷。可通過以下命令進行安裝:
pip install -U sentence-transformers
💻 使用示例
基礎用法
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)
📚 詳細文檔
🔍 評估結果
本模型使用 WikiAnswer 數據集中的測試集進行評估,評估結果如下:
[
{
"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
}
]
若需對本模型進行自動評估,請參考 Sentence Embeddings Benchmark:https://seb.sbert.net
🔧 訓練細節
本模型的訓練參數如下:
數據加載器 1
torch.utils.data.dataloader.DataLoader
,長度為 34123,參數如下:
{
"batch_size": 32,
"sampler": "torch.utils.data.sampler.RandomSampler",
"batch_sampler": "torch.utils.data.sampler.BatchSampler"
}
損失函數 1
sentence_transformers.losses.MultipleNegativesRankingLoss.MultipleNegativesRankingLoss
,參數如下:
{
"scale": 20.0,
"similarity_fct": "cos_sim"
}
數據加載器 2
torch.utils.data.dataloader.DataLoader
,長度為 51184,參數如下:
{
"batch_size": 32,
"sampler": "torch.utils.data.sampler.RandomSampler",
"batch_sampler": "torch.utils.data.sampler.BatchSampler"
}
損失函數 2
sentence_transformers.losses.OnlineContrastiveLoss.OnlineContrastiveLoss
fit() 方法的參數
{
"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
}
🔧 完整模型架構
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()
)
👥 貢獻者
感謝 @tradicio 添加此模型。
📄 許可證
本作品採用 知識共享署名 - 非商業性使用 - 相同方式共享 4.0 國際許可協議 進行許可。
