bobox's picture
batch_size = 64
04dc468 verified
metadata
language:
  - en
library_name: sentence-transformers
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:314315
  - loss:AdaptiveLayerLoss
  - loss:MultipleNegativesRankingLoss
base_model: microsoft/deberta-v3-small
datasets:
  - stanfordnlp/snli
  - sentence-transformers/stsb
metrics:
  - pearson_cosine
  - spearman_cosine
  - pearson_manhattan
  - spearman_manhattan
  - pearson_euclidean
  - spearman_euclidean
  - pearson_dot
  - spearman_dot
  - pearson_max
  - spearman_max
  - cosine_accuracy
  - cosine_accuracy_threshold
  - cosine_f1
  - cosine_f1_threshold
  - cosine_precision
  - cosine_recall
  - cosine_ap
  - dot_accuracy
  - dot_accuracy_threshold
  - dot_f1
  - dot_f1_threshold
  - dot_precision
  - dot_recall
  - dot_ap
  - manhattan_accuracy
  - manhattan_accuracy_threshold
  - manhattan_f1
  - manhattan_f1_threshold
  - manhattan_precision
  - manhattan_recall
  - manhattan_ap
  - euclidean_accuracy
  - euclidean_accuracy_threshold
  - euclidean_f1
  - euclidean_f1_threshold
  - euclidean_precision
  - euclidean_recall
  - euclidean_ap
  - max_accuracy
  - max_accuracy_threshold
  - max_f1
  - max_f1_threshold
  - max_precision
  - max_recall
  - max_ap
widget:
  - source_sentence: Two teenage girls conversing next to lockers.
    sentences:
      - Girls talking about their problems next to lockers.
      - A bully tries to pop a balloon without being caught in the act.
      - Two dogs standing together in the yard.
  - source_sentence: >-
      A young man in a heavy brown winter coat stands in front of a blue railing
      with his arms spread.
    sentences:
      - >-
        a boy holding onto the wall of an old brick house's raised foundation as
        construction occurs
      - The railing is in front of a frozen lake.
      - A skateboarder is doing tricks for a competition.
  - source_sentence: >-
      A shirtless man with a white hat and no shoes sitting crisscross with his
      back against the wall holding up a white plastic cup.
    sentences:
      - >-
        A long-haired boy riding his skateboard at a fast pace over a stone wall
        with graffiti.
      - A man is sitting crisscross
      - a child in a black ninja suit does a kick
  - source_sentence: A light colored dog leaps over a hurdle.
    sentences:
      - Men sit on the bus going to work,
      - A dog jumps over a obstacel.
      - a man standing on his motorbike.
  - source_sentence: people are standing near water with a boat heading their direction
    sentences:
      - >-
        People are standing near water with a large blue boat heading their
        direction.
      - Two people climbing on a wooden scaffold.
      - The dogs are near the toy.
pipeline_tag: sentence-similarity
model-index:
  - name: SentenceTransformer based on microsoft/deberta-v3-small
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: pearson_cosine
            value: 0.7660217567682521
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.7681125489633884
            name: Spearman Cosine
          - type: pearson_manhattan
            value: 0.7917532885619117
            name: Pearson Manhattan
          - type: spearman_manhattan
            value: 0.794675885405013
            name: Spearman Manhattan
          - type: pearson_euclidean
            value: 0.7860948725725584
            name: Pearson Euclidean
          - type: spearman_euclidean
            value: 0.7895594746178918
            name: Spearman Euclidean
          - type: pearson_dot
            value: 0.644843928972524
            name: Pearson Dot
          - type: spearman_dot
            value: 0.6427588138459626
            name: Spearman Dot
          - type: pearson_max
            value: 0.7917532885619117
            name: Pearson Max
          - type: spearman_max
            value: 0.794675885405013
            name: Spearman Max
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: cosine_accuracy
            value: 0.6730608840700584
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.5814725160598755
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.7170495061078964
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.4670722782611847
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.5977392321184954
            name: Cosine Precision
          - type: cosine_recall
            value: 0.895866802979407
            name: Cosine Recall
          - type: cosine_ap
            value: 0.7193483203625508
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.6444764576541057
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 71.95508575439453
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.7094262988661364
            name: Dot F1
          - type: dot_f1_threshold
            value: 53.77289581298828
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.5779411764705882
            name: Dot Precision
          - type: dot_recall
            value: 0.9183584051409376
            name: Dot Recall
          - type: dot_ap
            value: 0.6828334101602328
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.6664644779740693
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 213.6251678466797
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.7047102517243412
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 245.20578002929688
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.5908461842625544
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.8729370527238206
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.7132026586783923
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.6621426946698006
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 10.358880996704102
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.7024081560907013
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 12.010871887207031
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.5864970645792563
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.8754198919234701
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.7101786172295015
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.6730608840700584
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 213.6251678466797
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.7170495061078964
            name: Max F1
          - type: max_f1_threshold
            value: 245.20578002929688
            name: Max F1 Threshold
          - type: max_precision
            value: 0.5977392321184954
            name: Max Precision
          - type: max_recall
            value: 0.9183584051409376
            name: Max Recall
          - type: max_ap
            value: 0.7193483203625508
            name: Max Ap

SentenceTransformer based on microsoft/deberta-v3-small

This is a sentence-transformers model finetuned from microsoft/deberta-v3-small on the stanfordnlp/snli dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

  • Model Type: Sentence Transformer
  • Base model: microsoft/deberta-v3-small
  • Maximum Sequence Length: 512 tokens
  • Output Dimensionality: 768 tokens
  • Similarity Function: Cosine Similarity
  • Training Dataset:
  • Language: en

Model Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: DebertaV2Model 
  (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, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("bobox/DeBERTaV3-small-SentenceTransformer-AdaptiveLayerBaseline")
# Run inference
sentences = [
    'people are standing near water with a boat heading their direction',
    'People are standing near water with a large blue boat heading their direction.',
    'The dogs are near the toy.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Semantic Similarity

Metric Value
pearson_cosine 0.766
spearman_cosine 0.7681
pearson_manhattan 0.7918
spearman_manhattan 0.7947
pearson_euclidean 0.7861
spearman_euclidean 0.7896
pearson_dot 0.6448
spearman_dot 0.6428
pearson_max 0.7918
spearman_max 0.7947

Binary Classification

Metric Value
cosine_accuracy 0.6731
cosine_accuracy_threshold 0.5815
cosine_f1 0.717
cosine_f1_threshold 0.4671
cosine_precision 0.5977
cosine_recall 0.8959
cosine_ap 0.7193
dot_accuracy 0.6445
dot_accuracy_threshold 71.9551
dot_f1 0.7094
dot_f1_threshold 53.7729
dot_precision 0.5779
dot_recall 0.9184
dot_ap 0.6828
manhattan_accuracy 0.6665
manhattan_accuracy_threshold 213.6252
manhattan_f1 0.7047
manhattan_f1_threshold 245.2058
manhattan_precision 0.5908
manhattan_recall 0.8729
manhattan_ap 0.7132
euclidean_accuracy 0.6621
euclidean_accuracy_threshold 10.3589
euclidean_f1 0.7024
euclidean_f1_threshold 12.0109
euclidean_precision 0.5865
euclidean_recall 0.8754
euclidean_ap 0.7102
max_accuracy 0.6731
max_accuracy_threshold 213.6252
max_f1 0.717
max_f1_threshold 245.2058
max_precision 0.5977
max_recall 0.9184
max_ap 0.7193

Training Details

Training Dataset

stanfordnlp/snli

  • Dataset: stanfordnlp/snli at cdb5c3d
  • Size: 314,315 training samples
  • Columns: sentence1, sentence2, and label
  • Approximate statistics based on the first 1000 samples:
    sentence1 sentence2 label
    type string string int
    details
    • min: 5 tokens
    • mean: 16.62 tokens
    • max: 62 tokens
    • min: 4 tokens
    • mean: 9.46 tokens
    • max: 29 tokens
    • 0: 100.00%
  • Samples:
    sentence1 sentence2 label
    A person on a horse jumps over a broken down airplane. A person is outdoors, on a horse. 0
    Children smiling and waving at camera There are children present 0
    A boy is jumping on skateboard in the middle of a red bridge. The boy does a skateboarding trick. 0
  • Loss: AdaptiveLayerLoss with these parameters:
    {
        "loss": "MultipleNegativesRankingLoss",
        "n_layers_per_step": 1,
        "last_layer_weight": 1,
        "prior_layers_weight": 1,
        "kl_div_weight": 1.2,
        "kl_temperature": 1.2
    }
    

Evaluation Dataset

sentence-transformers/stsb

  • Dataset: sentence-transformers/stsb at ab7a5ac
  • Size: 1,500 evaluation samples
  • Columns: sentence1, sentence2, and score
  • Approximate statistics based on the first 1000 samples:
    sentence1 sentence2 score
    type string string float
    details
    • min: 5 tokens
    • mean: 14.77 tokens
    • max: 45 tokens
    • min: 6 tokens
    • mean: 14.74 tokens
    • max: 49 tokens
    • min: 0.0
    • mean: 0.47
    • max: 1.0
  • Samples:
    sentence1 sentence2 score
    A man with a hard hat is dancing. A man wearing a hard hat is dancing. 1.0
    A young child is riding a horse. A child is riding a horse. 0.95
    A man is feeding a mouse to a snake. The man is feeding a mouse to the snake. 1.0
  • Loss: AdaptiveLayerLoss with these parameters:
    {
        "loss": "MultipleNegativesRankingLoss",
        "n_layers_per_step": 1,
        "last_layer_weight": 1,
        "prior_layers_weight": 1,
        "kl_div_weight": 1.2,
        "kl_temperature": 1.2
    }
    

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • per_device_train_batch_size: 32
  • per_device_eval_batch_size: 16
  • learning_rate: 5e-06
  • weight_decay: 1e-07
  • num_train_epochs: 2
  • warmup_ratio: 0.5
  • save_safetensors: False
  • fp16: True
  • push_to_hub: True
  • hub_model_id: bobox/DeBERTaV3-small-SentenceTransformer-AdaptiveLayerBaselinen
  • hub_strategy: checkpoint
  • batch_sampler: no_duplicates

All Hyperparameters

Click to expand
  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: steps
  • prediction_loss_only: True
  • per_device_train_batch_size: 32
  • per_device_eval_batch_size: 16
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 1
  • eval_accumulation_steps: None
  • learning_rate: 5e-06
  • weight_decay: 1e-07
  • adam_beta1: 0.9
  • adam_beta2: 0.999
  • adam_epsilon: 1e-08
  • max_grad_norm: 1.0
  • num_train_epochs: 2
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.5
  • warmup_steps: 0
  • log_level: passive
  • log_level_replica: warning
  • log_on_each_node: True
  • logging_nan_inf_filter: True
  • save_safetensors: False
  • 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: True
  • 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: True
  • resume_from_checkpoint: None
  • hub_model_id: bobox/DeBERTaV3-small-SentenceTransformer-AdaptiveLayerBaselinen
  • hub_strategy: checkpoint
  • 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
  • batch_sampler: no_duplicates
  • multi_dataset_batch_sampler: proportional

Training Logs

Epoch Step Training Loss loss max_ap spearman_cosine
None 0 - 4.1425 - 0.4276
0.1001 983 4.7699 3.8387 0.6364 -
0.2001 1966 3.5997 2.7649 0.6722 -
0.3002 2949 2.811 2.3520 0.6838 -
0.4003 3932 2.414 2.0700 0.6883 -
0.5004 4915 2.186 1.8993 0.6913 -
0.6004 5898 1.8523 1.5632 0.7045 -
0.7005 6881 0.6415 1.4902 0.7082 -
0.8006 7864 0.5016 1.4636 0.7108 -
0.9006 8847 0.4194 1.3875 0.7121 -
1.0007 9830 0.3737 1.3077 0.7117 -
1.1008 10813 1.8087 1.0903 0.7172 -
1.2009 11796 1.6631 1.0388 0.7180 -
1.3009 12779 1.6161 1.0177 0.7169 -
1.4010 13762 1.5378 1.0136 0.7148 -
1.5011 14745 1.5215 1.0053 0.7159 -
1.6011 15728 1.2887 0.9600 0.7166 -
1.7012 16711 0.3058 0.9949 0.7180 -
1.8013 17694 0.2897 0.9792 0.7186 -
1.9014 18677 0.275 0.9598 0.7192 -
2.0 19646 - 0.9796 0.7193 -
None 0 - 2.4594 0.7193 0.7681

Framework Versions

  • Python: 3.10.13
  • Sentence Transformers: 3.0.1
  • Transformers: 4.41.2
  • PyTorch: 2.1.2
  • Accelerate: 0.30.1
  • Datasets: 2.19.2
  • Tokenizers: 0.19.1

Citation

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",
}

AdaptiveLayerLoss

@misc{li20242d,
    title={2D Matryoshka Sentence Embeddings}, 
    author={Xianming Li and Zongxi Li and Jing Li and Haoran Xie and Qing Li},
    year={2024},
    eprint={2402.14776},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply}, 
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}