Edit model card

SentenceTransformer based on sentence-transformers/multi-qa-MiniLM-L6-cos-v1

This is a sentence-transformers model finetuned from sentence-transformers/multi-qa-MiniLM-L6-cos-v1. It maps sentences & paragraphs to a 384-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 Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, '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()
)

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("Trelis/multi-qa-MiniLM-L6-cos-v1-ft-pairs")
# Run inference
sentences = [
    'What is the ruling if an attacking player intentionally passes the ball to a defending player seeking a rebound or restart of the touch count?',
    'ball goes to ground, a change of possession occurs. 12. 6 if a player from the defending team unintentionally makes contact with the ball in flight and the ball is retrieved by an attacking player, play and the touch count continues. 12. 7 a player from the attacking team cannot pass the ball into a defending player intentionally seeking a rebound or a restart of the touch count. ruling = a penalty to the defending team at the point of the pass. 13 the rollball 13. 1 the attacking player is to position on the mark, face the opponent ’ s try line, make a genuine attempt to stand parallel to the sidelines, place the ball on the ground between the feet in a controlled manner and : 13. 1. 1 step forward over the ball ; or 13. 1. 2 roll the ball back between the feet no more than one ( 1 ) metre ; or 13. 1. 3 pass a foot over the ball. ruling = a change of possession to the defending team at the point of the infringement. 13. 2 a player must perform the rollball on the mark. ruling = a penalty to the defending team at the point of the infringement. 13. 3 a player must not perform a voluntary rollball',
    '. see appendix 1. kick strike or propel forcibly with the foot, a blow or forceful thrust with the foot to the ball. a tap to commence or recommence play or a penalty tap is not defined as a kick. line markings markings of the field of play. see appendix 1. link the player beside the wing player. mark ( for a tap ) the centre of the halfway line for the commencement or recommencement of play, or the position where a penalty tap is awarded as a result of an infringement. mark ( for a touch ) the position in the field of play where the player in possession was at the time the touch was made. fit playing rules - 5th edition 2 copyright © touch football australia 2020 middle the player inside the link player. nta national touch association as defined in the fit constitution. obstruction a deliberate attempt by either an attacking or defending player to gain an unfair advantage by interfering with the opposition to prevent them from gaining a rightful advantage. offside ( attacker ) an attacking player in a position forward of the ball. offside ( defender ) a defending player in a position closer than seven ( 7 ) metres from the mark of the rollball ; or ten ( 10 ) metres from the mark',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

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

Training Details

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • per_device_train_batch_size: 16
  • per_device_eval_batch_size: 16
  • learning_rate: 0.0001
  • num_train_epochs: 2
  • lr_scheduler_type: cosine
  • warmup_ratio: 0.3

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
  • learning_rate: 0.0001
  • 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: cosine
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.3
  • 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
  • batch_sampler: batch_sampler
  • multi_dataset_batch_sampler: proportional

Training Logs

Epoch Step Training Loss loss
0.1429 2 1.0897 -
0.2857 4 1.3506 1.1206
0.4286 6 1.1968 -
0.5714 8 1.4074 1.0205
0.7143 10 1.3476 -
0.8571 12 1.0062 1.0278
1.0 14 1.4792 -
1.1429 16 0.8863 1.1568
1.2857 18 0.5465 -
1.4286 20 0.5672 1.1830
1.5714 22 0.5482 -
1.7143 24 0.7633 1.1838
1.8571 26 0.5931 -
2.0 28 0.4969 1.1800

Framework Versions

  • Python: 3.10.12
  • Sentence Transformers: 3.0.1
  • Transformers: 4.42.3
  • PyTorch: 2.1.1+cu121
  • Accelerate: 0.31.0
  • Datasets: 2.17.1
  • 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",
}

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}
}
Downloads last month
8
Safetensors
Model size
22.7M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for Trelis/multi-qa-MiniLM-L6-cos-v1-ft-pairs

Finetuned
(9)
this model