ONNX 模型库
返回模型

说明文档

基于 sentence-transformers/all-MiniLM-L6-v2 的 SentenceTransformer

这是一个基于 sentence-transformers/all-MiniLM-L6-v2 微调的 sentence-transformers 模型。它将句子和段落映射到 384 维的稠密向量空间,可用于语义文本相似度、语义搜索、释义挖掘、文本分类、聚类等任务。

模型详情

模型描述

  • 模型类型: Sentence Transformer
  • 基础模型: sentence-transformers/all-MiniLM-L6-v2 <!-- at revision fa97f6e7cb1a59073dff9e6b13e2715cf7475ac9 -->
  • 最大序列长度: 256 个 token
  • 输出维度: 384 维
  • 相似度函数: 余弦相似度 <!-- - 训练数据集: 未知 --> <!-- - 语言: 未知 --> <!-- - 许可证: 未知 -->

模型来源

完整模型架构

SentenceTransformer(
  (0): Transformer({'max_seq_length': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

使用方法

直接使用 (Sentence Transformers)

首先安装 Sentence Transformers 库:

pip install -U sentence-transformers

然后你可以加载此模型并运行推理。

from sentence_transformers import SentenceTransformer

# 从 🤗 Hub 下载
model = SentenceTransformer("overfitting-co/A2P-constrastive-all")
# 运行推理
sentences = [
    'Khaosan Road',
    'Reserved',
    'Adventurous',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# 获取嵌入的相似度分数
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

<!--

直接使用 (Transformers)

<details><summary>点击查看 Transformers 中的直接用法</summary>

</details> -->

<!--

下游使用 (Sentence Transformers)

你可以在自己的数据集上微调此模型。

<details><summary>点击展开</summary>

</details> -->

<!--

超出范围的使用

列出模型可能被预见性地滥用的方式,并说明用户不应对模型做什么。 -->

评估

指标

二分类

指标 数值
cosine_accuracy 0.9574
cosine_accuracy_threshold 0.8163
cosine_f1 0.958
cosine_f1_threshold 0.8131
cosine_precision 0.9682
cosine_recall 0.9481
cosine_ap 0.9909
dot_accuracy 0.9574
dot_accuracy_threshold 0.8163
dot_f1 0.958
dot_f1_threshold 0.8131
dot_precision 0.9682
dot_recall 0.9481
dot_ap 0.9909
manhattan_accuracy 0.9609
manhattan_accuracy_threshold 9.5648
manhattan_f1 0.9619
manhattan_f1_threshold 9.5648
manhattan_precision 0.9619
manhattan_recall 0.9619
manhattan_ap 0.9909
euclidean_accuracy 0.9574
euclidean_accuracy_threshold 0.6061
euclidean_f1 0.958
euclidean_f1_threshold 0.6114
euclidean_precision 0.9682
euclidean_recall 0.9481
euclidean_ap 0.9909
max_accuracy 0.9609
max_accuracy_threshold 9.5648
max_f1 0.9619
max_f1_threshold 9.5648
max_precision 0.9682
max_recall 0.9619
max_ap 0.9909

二分类

指标 数值
cosine_accuracy 0.9592
cosine_accuracy_threshold 0.7969
cosine_f1 0.9591
cosine_f1_threshold 0.7969
cosine_precision 0.9574
cosine_recall 0.9609
cosine_ap 0.9878
dot_accuracy 0.9592
dot_accuracy_threshold 0.7969
dot_f1 0.9591
dot_f1_threshold 0.7969
dot_precision 0.9574
dot_recall 0.9609
dot_ap 0.9878
manhattan_accuracy 0.9557
manhattan_accuracy_threshold 9.8085
manhattan_f1 0.9558
manhattan_f1_threshold 9.917
manhattan_precision 0.9507
manhattan_recall 0.9609
manhattan_ap 0.9866
euclidean_accuracy 0.9592
euclidean_accuracy_threshold 0.6373
euclidean_f1 0.9591
euclidean_f1_threshold 0.6373
euclidean_precision 0.9574
euclidean_recall 0.9609
euclidean_ap 0.9878
max_accuracy 0.9592
max_accuracy_threshold 9.8085
max_f1 0.9591
max_f1_threshold 9.917
max_precision 0.9574
max_recall 0.9609
max_ap 0.9878

<!--

偏见、风险和局限性

该模型存在哪些已知或可预见的问题?你也可以在这里标记已知的失败案例或模型的弱点。 -->

<!--

建议

针对可预见的问题有什么建议?例如,过滤显式内容。 -->

训练详情

训练数据集

未命名数据集

  • 规模:4,505 个训练样本
  • 列:<code>sentence_0</code>、<code>sentence_1</code> 和 <code>label</code>
  • 基于前 1000 个样本的近似统计:
    sentence_0 sentence_1 label
    类型 字符串 字符串 整数
    详情 <ul><li>最小值:3 个 token</li><li>平均值:6.49 个 token</li><li>最大值:24 个 token</li></ul> <ul><li>最小值:3 个 token</li><li>平均值:3.79 个 token</li><li>最大值:8 个 token</li></ul> <ul><li>0:约 52.30%</li><li>1:约 47.70%</li></ul>
  • 样本:
    sentence_0 sentence_1 label
    <code>N Seoul Tower</code> <code>Laid-back</code> <code>0</code>
    <code>Magere Brug</code> <code>Romantic</code> <code>1</code>
    <code>Polynesian Cultural Center</code> <code>Adventurous</code> <code>1</code>
  • 损失函数:<code>OnlineContrastiveLoss</code>

训练超参数

非默认超参数

  • per_device_train_batch_size: 32
  • per_device_eval_batch_size: 32
  • num_train_epochs: 5
  • multi_dataset_batch_sampler: round_robin

所有超参数

<details><summary>点击展开</summary>

  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: no
  • prediction_loss_only: True
  • per_device_train_batch_size: 32
  • per_device_eval_batch_size: 32
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 1
  • eval_accumulation_steps: None
  • torch_empty_cache_steps: None
  • learning_rate: 5e-05
  • weight_decay: 0.0
  • adam_beta1: 0.9
  • adam_beta2: 0.999
  • adam_epsilon: 1e-08
  • max_grad_norm: 1
  • num_train_epochs: 5
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.0
  • warmup_steps: 0
  • log_level: passive
  • log_level_replica: warning
  • log_on_each_node: True
  • logging_nan_inf_filter: True
  • save_safetensors: True
  • save_on_each_node: False
  • save_only_model: False
  • restore_callback_states_from_checkpoint: False
  • no_cuda: False
  • use_cpu: False
  • use_mps_device: False
  • seed: 42
  • data_seed: None
  • jit_mode_eval: False
  • use_ipex: False
  • bf16: False
  • fp16: False
  • fp16_opt_level: O1
  • half_precision_backend: auto
  • bf16_full_eval: False
  • fp16_full_eval: False
  • tf32: None
  • local_rank: 0
  • ddp_backend: None
  • tpu_num_cores: None
  • tpu_metrics_debug: False
  • debug: []
  • dataloader_drop_last: False
  • dataloader_num_workers: 0
  • dataloader_prefetch_factor: None
  • past_index: -1
  • disable_tqdm: False
  • remove_unused_columns: True
  • label_names: None
  • load_best_model_at_end: False
  • ignore_data_skip: False
  • fsdp: []
  • fsdp_min_num_params: 0
  • fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
  • fsdp_transformer_layer_cls_to_wrap: None
  • accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
  • deepspeed: None
  • label_smoothing_factor: 0.0
  • optim: adamw_torch
  • optim_args: None
  • adafactor: False
  • group_by_length: False
  • length_column_name: length
  • ddp_find_unused_parameters: None
  • ddp_bucket_cap_mb: None
  • ddp_broadcast_buffers: False
  • dataloader_pin_memory: True
  • dataloader_persistent_workers: False
  • skip_memory_metrics: True
  • use_legacy_prediction_loop: False
  • push_to_hub: False
  • resume_from_checkpoint: None
  • hub_model_id: None
  • hub_strategy: every_save
  • hub_private_repo: False
  • hub_always_push: False
  • gradient_checkpointing: False
  • gradient_checkpointing_kwargs: None
  • include_inputs_for_metrics: False
  • eval_do_concat_batches: True
  • fp16_backend: auto
  • push_to_hub_model_id: None
  • push_to_hub_organization: None
  • mp_parameters:
  • auto_find_batch_size: False
  • full_determinism: False
  • torchdynamo: None
  • ray_scope: last
  • ddp_timeout: 1800
  • torch_compile: False
  • torch_compile_backend: None
  • torch_compile_mode: None
  • dispatch_batches: None
  • split_batches: None
  • include_tokens_per_second: False
  • include_num_input_tokens_seen: False
  • neftune_noise_alpha: None
  • optim_target_modules: None
  • batch_eval_metrics: False
  • eval_on_start: False
  • eval_use_gather_object: False
  • batch_sampler: batch_sampler
  • multi_dataset_batch_sampler: round_robin

</details>

训练日志

轮次 步数 训练损失 max_ap test_max_ap
1.0 141 - 0.6780 -
2.0 282 - 0.7538 -
3.0 423 - 0.8064 -
3.5461 500 6.7404 - -
4.0 564 - 0.9751 -
5.0 705 - 0.9909 0.9878

框架版本

  • Python: 3.10.12
  • Sentence Transformers: 3.2.1
  • Transformers: 4.44.2
  • PyTorch: 2.5.0+cu121
  • Accelerate: 0.34.2
  • Datasets: 3.1.0
  • Tokenizers: 0.19.1

引用

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

<!--

术语表

明确定义术语,以便各领域的读者都能理解。 -->

<!--

模型卡片作者

列出创建模型卡片的人员,为模型卡片构建中的详细工作提供认可和责任归属。 -->

<!--

模型卡片联系方式

为有模型卡片更新、建议或问题的人提供联系模型卡片作者的方式。 -->

overfitting-co/A2P-constrastive-all

作者 overfitting-co

sentence-similarity sentence-transformers
↓ 1 ♥ 0

创建时间: 2024-11-05 17:32:17+00:00

更新时间: 2024-11-05 18:15:32+00:00

在 Hugging Face 上查看

文件 (21)

.gitattributes
1_Pooling/config.json
README.md
config.json
config_sentence_transformers.json
model.safetensors
modules.json
onnx/model.onnx ONNX
repository/onnx/config.json
repository/onnx/model.onnx ONNX
repository/onnx/special_tokens_map.json
repository/onnx/test
repository/onnx/test.py
repository/onnx/tokenizer.json
repository/onnx/tokenizer_config.json
repository/onnx/vocab.txt
sentence_bert_config.json
special_tokens_map.json
tokenizer.json
tokenizer_config.json
vocab.txt