This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L12-v2. 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 Type: Sentence Transformer
Base model: sentence-transformers/all-MiniLM-L12-v2
Maximum Sequence Length: 128 tokens
Output Dimensionality: 384 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': 128, '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("Trelis/all-MiniLM-L12-v2-ft-pairs-balanced")
# Run inference
sentences = [
    'What happens if a player deliberately delays the changeover procedure after a Change of Possession?',
    ' Registration\n5\n03 I\nThe Ball\n6\n04 I\nPlaying Uniform\n6\n05 I\nTeam Composition\n6\n06 I\nTeam Coach and Team Officials\n7\n07\nI\nCommencement and Recommencement of Play\n7\n08\nI\nMatch Duration\n8\n09 I\nPossession\n8\n10\nI\nThe Touch\n9\n11\nI\nPassing\n10\n12\nI\nBall Touched in Flight\n10\n13\nI\nThe Rollball\n11\n14\nI\nScoring\n13\n15\nI\nOffside\n13\n16\nI\nObstruction\n14\n17\nI\nInterchange\n14\n18\nI\nPenalty\n15\n19\nI\nAdvantage\n16\n20\nI\nMisconduct\n16\n21\nI\nForced Interchange\n16\n22\nI\nSin Bin\n16\n23\nI\nDismissal\n17\n24\nI\nDrop-Off\n17\n25\nI\nMatch Officials\n18\nFIT Playing Rules - 5th Edition\nCOPYRIGHT © Touch Football Australia 2020\nFIT Playing Rules - 5th Edition\nCOPYRIGHT © Touch Football Australia 2020\n  Definitions and Terminology  \nUnless the contrary intention appears, the following deﬁnitions and terminology apply \nto the game of Touch:\nTERM/PHRASE\nDEFINITION/DESCRIPTION\nAdvantage\nThe period of time after an Infringement in which the non-offending \nside has the opportunity to gain Advantage either territorial, tactical \nor in the form of a Try.\nAttacking Try Line\nThe line on or over which a player has to place the ball to \nscore a Try.\nAttacking Team\nThe Team which has or is gaining Possession.\nBehind\nA position or direction towards a Team’s Defending Try Line.\nChange of Possession\nThe act of moving control of the ball from one Team to the other.\nDead/Dead Ball\nWhen the ball is out of play including the period following a Try and \nuntil the match is recommenced and when the ball goes to ground \nand/or outside the boundaries of the Field of Play prior to the \nsubsequent Rollball.\nDead Ball Line\nThe end boundaries of the Field of Play. There is one at each end of \nthe Field of Play. See Appendix 1.\nDef',
    ' Registration\n5\n03 I\nThe Ball\n6\n04 I\nPlaying Uniform\n6\n05 I\nTeam Composition\n6\n06 I\nTeam Coach and Team Officials\n7\n07\nI\nCommencement and Recommencement of Play\n7\n08\nI\nMatch Duration\n8\n09 I\nPossession\n8\n10\nI\nThe Touch\n9\n11\nI\nPassing\n10\n12\nI\nBall Touched in Flight\n10\n13\nI\nThe Rollball\n11\n14\nI\nScoring\n13\n15\nI\nOffside\n13\n16\nI\nObstruction\n14\n17\nI\nInterchange\n14\n18\nI\nPenalty\n15\n19\nI\nAdvantage\n16\n20\nI\nMisconduct\n16\n21\nI\nForced Interchange\n16\n22\nI\nSin Bin\n16\n23\nI\nDismissal\n17\n24\nI\nDrop-Off\n17\n25\nI\nMatch Officials\n18\nFIT Playing Rules - 5th Edition\nCOPYRIGHT © Touch Football Australia 2020\nFIT Playing Rules - 5th Edition\nCOPYRIGHT © Touch Football Australia 2020\n  Definitions and Terminology  \nUnless the contrary intention appears, the following deﬁnitions and terminology apply \nto the game of Touch:\nTERM/PHRASE\nDEFINITION/DESCRIPTION\nAdvantage\nThe period of time after an Infringement in which the non-offending \nside has the opportunity to gain Advantage either territorial, tactical \nor in the form of a Try.\nAttacking Try Line\nThe line on or over which a player has to place the ball to \nscore a Try.\nAttacking Team\nThe Team which has or is gaining Possession.\nBehind\nA position or direction towards a Team’s Defending Try Line.\nChange of Possession\nThe act of moving control of the ball from one Team to the other.\nDead/Dead Ball\nWhen the ball is out of play including the period following a Try and \nuntil the match is recommenced and when the ball goes to ground \nand/or outside the boundaries of the Field of Play prior to the \nsubsequent Rollball.\nDead Ball Line\nThe end boundaries of the Field of Play. There is one at each end of \nthe Field of Play. See Appendix 1.\nDef',
]
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: 4
per_device_eval_batch_size: 4
learning_rate: 1e-05
num_train_epochs: 1
lr_scheduler_type: cosine
warmup_ratio: 0.3
bf16: 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: 4
per_device_eval_batch_size: 4
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
learning_rate: 1e-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: 1
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: True
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
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss
0.1053	2	4.6868	-
0.1579	3	-	2.7075
0.2105	4	5.703	-
0.3158	6	2.1691	2.6412
0.4211	8	1.705	-
0.4737	9	-	2.6254
0.5263	10	1.7985	-
0.6316	12	3.4822	2.6087
0.7368	14	4.2724	-
0.7895	15	-	2.6000
0.8421	16	3.1489	-
0.9474	18	5.7594	2.6032

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.41.2
PyTorch: 2.1.1+cu121
Accelerate: 0.31.0
Datasets: 2.17.1
Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

bibtex

@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",
}

CoSENTLoss

bibtex

@online{kexuefm-8847,
    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
    author={Su Jianlin},
    year={2022},
    month={Jan},
    url={https://kexue.fm/archives/8847},
}