metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:6233
  - loss:OnlineContrastiveLoss
base_model: sentence-transformers/all-mpnet-base-v2
widget:
  - source_sentence: >-
      as permitted by applicable law , in no event shall groupon , its
      subsidiaries or affiliates or any of their respective employees , officers
      , directors , agents , merchants , partners , third-party content
      providers or licensors , or any of their officers , directors , employees
      , or agents , be liable for any direct or indirect lost profits or lost
      business damages , indirect , incidental , special , consequential , or
      punitive damages arising out of , related to , or in connection with any
      of the following : -lrb- a -rrb- your use of the site , the content , user
      content , including , without limitation , any personal information , and
      any other information either contained in the site or submitted by you to
      the site ; -lrb- b -rrb- your inability to use the site ; -lrb- c -rrb-
      modification or removal of content submitted on the site ; -lrb- d -rrb-
      the merchant offerings , products , and other available programs
      accessible or available through the site ; -lrb- e -rrb- any products or
      services purchased or obtained directly from a merchant ; -lrb- f -rrb-
      these terms of use ; or -lrb- g -rrb- any improper use of information you
      provide to the site , including , without limitation , any personal
      information .
    sentences:
      - >-
        since the clause states that the provider is not liable for any loss
        resulting from the use of the service and or of the website, including
        lost profits, lost opportunity, lost business or lost sales
      - >-
        since the clause states that the provider is not liable for any special,
        direct and/or indirect, punitive, incidental or consequential  damage,
        including negligence, harm or failure
      - >-
        since the contract or access may be terminated where the user fails to
        maintain a prescribed level of reputation.
  - source_sentence: >-
      however , vivino reserves the right to -lrb- i -rrb- remove , suspend ,
      edit or modify any content in its sole discretion , including without
      limitation any user submissions at any time , without notice to you and
      for any reason -lrb- including , but not limited to , upon receipt of
      claims or allegations from third parties or authorities relating to such
      content or if vivino is concerned that you may have violated these terms
      of use -rrb- , or for no reason at all and -lrb- ii -rrb- to remove ,
      suspend or block any user submissions from the service .
    sentences:
      - >-
        Since the clause states that the provider has the right to remove
        content and material if they constitute a violation of third party
        rights, including trademarks
      - >-
        since the clause states that except as required by law, or to the
        fullest extent permissible by applicable law the provider is not liable,
        or that the users are solely responsible for ensuring that the Terms of
        Use/Service are in compliance with all laws, rules and regulations 
      - >-
        since the clause states that the compensation for liability or aggregate
        liability is limited to, or should not exceed, a certain total amount,
        or that the sole remedy is to stop using the service and cancel the
        account, or that you can't recover any  damages or losses
  - source_sentence: >-
      we will not incur any liability or responsibility if we choose to remove ,
      disable or delete such access or ability to use any or all portion -lrb- s
      -rrb- of the services .
    sentences:
      - >-
        since the clause states that except as required by law, or to the
        fullest extent permissible by applicable law the provider is not liable,
        or that the users are solely responsible for ensuring that the Terms of
        Use/Service are in compliance with all laws, rules and regulations 
      - >-
        since the clause states that the provider is not liable under different
        theories of liability, including tort law, contract law,  strict
        liability, statutory liability, product liability and other liability
        theories
      - >-
        since the clause mentions the contract or access may be terminated but
        does not state the grounds for termination.
  - source_sentence: >-
      in such event , supercell shall not be required to provide refunds ,
      benefits or other compensation to users in connection with such
      discontinued service .
    sentences:
      - >-
        since the clause states that the provider is not liable even if he was,
        or should have been, aware or have been advised about the possibility of
        any damage or loss
      - >-
        since the contract or access can be terminated where the user fails to
        adhere to its terms, or community standards, or the spirit of the ToS or
        community terms, including inappropriate behaviour, using cheats or
        other disallowed practices to improve their situation in the service,
        deriving disallowed profits from the service, or interfering with other
        users' enjoyment of the service or otherwise puts them at risk, or is
        investigated under any suspision of misconduct.
      - >-
        since the clause states that the provider is not liable for any
        technical problems, failure, suspension, disruption, modification,
        discontinuance, unavailability of service, any unilateral change,
        unilateral termination,  unilateral limitation  including  limits on
        certain features and services or restricttion to  access to parts or all
        of the Service without notice 
  - source_sentence: >-
      we may change the price of the services at any time and if you have a
      recurring purchase , we will notify you by email at least 15 days before
      the price change .
    sentences:
      - >-
        Since the clause states that the provider has the right for unilateral
        change of the contract/services/goods/features for any reason at its
        full discretion, at any time 
      - >-
        Since the clause states that the provider has the right for unilateral
        change of the contract/services/goods/features for any reason at its
        full discretion, at any time 
      - >-
        since the clause states that the provider is not liable even if he was,
        or should have been, aware or have been advised about the possibility of
        any damage or loss
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - 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
model-index:
  - name: SentenceTransformer based on sentence-transformers/all-mpnet-base-v2
    results:
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: eval
          type: eval
        metrics:
          - type: cosine_accuracy
            value: 0.8888888888888888
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.7393813133239746
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.8966442953020134
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.7284817099571228
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.8608247422680413
            name: Cosine Precision
          - type: cosine_recall
            value: 0.9355742296918768
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9472776717150163
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.8888888888888888
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.7393813133239746
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.8966442953020134
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.7284817099571228
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.8608247422680413
            name: Dot Precision
          - type: dot_recall
            value: 0.9355742296918768
            name: Dot Recall
          - type: dot_ap
            value: 0.9472776717150163
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.8888888888888888
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 15.613447189331055
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.896921017402945
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 15.90174674987793
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.8589743589743589
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.938375350140056
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.947924181751851
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.8888888888888888
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.7219676971435547
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.8966442953020134
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.7369099855422974
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.8608247422680413
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.9355742296918768
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9472776717150163
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.8888888888888888
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 15.613447189331055
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.896921017402945
            name: Max F1
          - type: max_f1_threshold
            value: 15.90174674987793
            name: Max F1 Threshold
          - type: max_precision
            value: 0.8608247422680413
            name: Max Precision
          - type: max_recall
            value: 0.938375350140056
            name: Max Recall
          - type: max_ap
            value: 0.947924181751851
            name: Max Ap

SentenceTransformer based on sentence-transformers/all-mpnet-base-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-mpnet-base-v2. 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: sentence-transformers/all-mpnet-base-v2
Maximum Sequence Length: 384 tokens
Output Dimensionality: 768 tokens
Similarity Function: Cosine Similarity

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

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, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

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("cruzlorite/all-mpnet-base-v2-unfair-tos-rationale")
# Run inference
sentences = [
    'we may change the price of the services at any time and if you have a recurring purchase , we will notify you by email at least 15 days before the price change .',
    'Since the clause states that the provider has the right for unilateral change of the contract/services/goods/features for any reason at its full discretion, at any time ',
    'Since the clause states that the provider has the right for unilateral change of the contract/services/goods/features for any reason at its full discretion, at any time ',
]
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

Binary Classification

Dataset: eval
Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.8889
cosine_accuracy_threshold	0.7394
cosine_f1	0.8966
cosine_f1_threshold	0.7285
cosine_precision	0.8608
cosine_recall	0.9356
cosine_ap	0.9473
dot_accuracy	0.8889
dot_accuracy_threshold	0.7394
dot_f1	0.8966
dot_f1_threshold	0.7285
dot_precision	0.8608
dot_recall	0.9356
dot_ap	0.9473
manhattan_accuracy	0.8889
manhattan_accuracy_threshold	15.6134
manhattan_f1	0.8969
manhattan_f1_threshold	15.9017
manhattan_precision	0.859
manhattan_recall	0.9384
manhattan_ap	0.9479
euclidean_accuracy	0.8889
euclidean_accuracy_threshold	0.722
euclidean_f1	0.8966
euclidean_f1_threshold	0.7369
euclidean_precision	0.8608
euclidean_recall	0.9356
euclidean_ap	0.9473
max_accuracy	0.8889
max_accuracy_threshold	15.6134
max_f1	0.8969
max_f1_threshold	15.9017
max_precision	0.8608
max_recall	0.9384
max_ap	0.9479

Training Details

Training Dataset

Unnamed Dataset

Size: 6,233 training samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 label
type string string int
details
min: 8 tokens
mean: 63.0 tokens
max: 384 tokens

min: 10 tokens
mean: 41.12 tokens
max: 96 tokens

0: ~48.70%
1: ~51.30%

	sentence1	sentence2	label
type	string	string	int
details	min: 8 tokens mean: 63.0 tokens max: 384 tokens	min: 10 tokens mean: 41.12 tokens max: 96 tokens	0: ~48.70% 1: ~51.30%

Samples:

sentence1	sentence2	label
`we may revise these terms from time to time and the most current version will always be posted on our website .`	`Since the clause states that the provider has the right for unilateral change of the contract/services/goods/features where the notification of changes is left at a full discretion of the provider such as by simply posting the new terms on their website without a notification to the consumer`	`1`
neither fitbit , its suppliers , or licensors , nor any other party involved in creating , producing , or delivering the fitbit service will be liable for any incidental , special , exemplary , or consequential damages , including lost profits , loss of data or goodwill , service interruption , computer damage , or system failure or the cost of substitute services arising out of or in connection with these terms or from the use of or inability to use the fitbit service , whether based on warranty , contract , tort -lrb- including negligence -rrb- , product liability , or any other legal theory , and whether or not fitbit has been informed of the possibility of such damage , even if a limited remedy set forth herein is found to have failed of its essential purpose .	`since the clause states that the provider is not liable even if he was, or should have been, aware or have been advised about the possibility of any damage or loss`	`1`
`the company reserves the right -lrb- but has no obligation -rrb- , at its sole discretion and without prior notice to :`	`Since the clause states that the provider has the right to remove content and material if he believes that there is a case violation of terms such as acount tranfer, policies, standard, code of conduct`	`1`

Loss: OnlineContrastiveLoss

Evaluation Dataset

Unnamed Dataset

Size: 693 evaluation samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 693 samples:
sentence1 sentence2 label
type string string int
details
min: 8 tokens
mean: 63.59 tokens
max: 384 tokens

min: 10 tokens
mean: 42.75 tokens
max: 96 tokens

0: ~48.48%
1: ~51.52%

	sentence1	sentence2	label
type	string	string	int
details	min: 8 tokens mean: 63.59 tokens max: 384 tokens	min: 10 tokens mean: 42.75 tokens max: 96 tokens	0: ~48.48% 1: ~51.52%

Samples:

sentence1	sentence2	label
you expressly understand and agree that evernote , its subsidiaries , affiliates , service providers , and licensors , and our and their respective officers , employees , agents and successors shall not be liable to you for any direct , indirect , incidental , special , consequential or exemplary damages , including but not limited to , damages for loss of profits , goodwill , use , data , cover or other intangible losses -lrb- even if evernote has been advised of the possibility of such damages -rrb- resulting from : -lrb- i -rrb- the use or the inability to use the service or to use promotional codes or evernote points ; -lrb- ii -rrb- the cost of procurement of substitute services resulting from any data , information or service purchased or obtained or messages received or transactions entered into through or from the service ; -lrb- iii -rrb- unauthorized access to or the loss , corruption or alteration of your transmissions , content or data ; -lrb- iv -rrb- statements or conduct of any third party on or using the service , or providing any services related to the operation of the service ; -lrb- v -rrb- evernote 's actions or omissions in reliance upon your basic subscriber information and any changes thereto or notices received therefrom ; -lrb- vi -rrb- your failure to protect the confidentiality of any passwords or access rights to your account ; -lrb- vii -rrb- the acts or omissions of any third party using or integrating with the service ; -lrb- viii -rrb- any advertising content or your purchase or use of any advertised or other third-party product or service ; -lrb- ix -rrb- the termination of your account in accordance with the terms of these terms of service ; or -lrb- x -rrb- any other matter relating to the service .	`since the clause states that the provider is not liable for any information stored or processed within the Services, inaccuracies or error of information, content and material posted, software, products and services on the website, including copyright violation, defamation, slander, libel, falsehoods, obscenity, pornography, profanity, or objectionable material`	`1`
`to the fullest extent permitted by law , badoo expressly excludes :`	`since the clause states that the provider is not liable even if he was, or should have been, aware or have been advised about the possibility of any damage or loss`	`1`
`notwithstanding any other remedies available to truecaller , you agree that truecaller may suspend or terminate your use of the services without notice if you use the services or the content in any prohibited manner , and that such use will be deemed a material breach of these terms .`	`since the clause generally states the contract or access may be terminated in an event of a force majeure, act of God or other unforeseen events of a similar nature.`	`0`

Loss: OnlineContrastiveLoss

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
learning_rate: 2e-05
num_train_epochs: 2
warmup_ratio: 0.1
fp16: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 16
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
torch_empty_cache_steps: None
learning_rate: 2e-05
weight_decay: 0.0
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: 0.1
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: 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: 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
use_liger_kernel: False
eval_use_gather_object: False
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	eval_max_ap
0	0	-	-	0.6125
0.2564	100	0.9286	0.4118	0.8794
0.5128	200	0.3916	0.2868	0.9177
0.7692	300	0.3414	0.2412	0.9448
1.0256	400	0.2755	0.2103	0.9470
1.2821	500	0.1893	0.1892	0.9486
1.5385	600	0.1557	0.1709	0.9548
1.7949	700	0.1566	0.1888	0.9479

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.1.1
Transformers: 4.45.2
PyTorch: 2.5.1+cu121
Accelerate: 1.1.1
Datasets: 3.1.0
Tokenizers: 0.20.3

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",
}