This is a sentence-transformers model trained. It maps sentences & paragraphs to a 1024-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

Maximum Sequence Length: 8192 tokens
Output Dimensionality: 1024 dimensions
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(
  (transformer): Transformer(
    (auto_model): XLMRobertaLoRA(
      (roberta): XLMRobertaModel(
        (embeddings): XLMRobertaEmbeddings(
          (word_embeddings): ParametrizedEmbedding(
            250002, 1024, padding_idx=1
            (parametrizations): ModuleDict(
              (weight): ParametrizationList(
                (0): LoRAParametrization()
              )
            )
          )
          (token_type_embeddings): ParametrizedEmbedding(
            1, 1024
            (parametrizations): ModuleDict(
              (weight): ParametrizationList(
                (0): LoRAParametrization()
              )
            )
          )
        )
        (emb_drop): Dropout(p=0.1, inplace=False)
        (emb_ln): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (encoder): XLMRobertaEncoder(
          (layers): ModuleList(
            (0-23): 24 x Block(
              (mixer): MHA(
                (rotary_emb): RotaryEmbedding()
                (Wqkv): ParametrizedLinearResidual(
                  in_features=1024, out_features=3072, bias=True
                  (parametrizations): ModuleDict(
                    (weight): ParametrizationList(
                      (0): LoRAParametrization()
                    )
                  )
                )
                (inner_attn): FlashSelfAttention(
                  (drop): Dropout(p=0.1, inplace=False)
                )
                (inner_cross_attn): FlashCrossAttention(
                  (drop): Dropout(p=0.1, inplace=False)
                )
                (out_proj): ParametrizedLinear(
                  in_features=1024, out_features=1024, bias=True
                  (parametrizations): ModuleDict(
                    (weight): ParametrizationList(
                      (0): LoRAParametrization()
                    )
                  )
                )
              )
              (dropout1): Dropout(p=0.1, inplace=False)
              (drop_path1): StochasticDepth(p=0.0, mode=row)
              (norm1): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (mlp): Mlp(
                (fc1): ParametrizedLinear(
                  in_features=1024, out_features=4096, bias=True
                  (parametrizations): ModuleDict(
                    (weight): ParametrizationList(
                      (0): LoRAParametrization()
                    )
                  )
                )
                (fc2): ParametrizedLinear(
                  in_features=4096, out_features=1024, bias=True
                  (parametrizations): ModuleDict(
                    (weight): ParametrizationList(
                      (0): LoRAParametrization()
                    )
                  )
                )
              )
              (dropout2): Dropout(p=0.1, inplace=False)
              (drop_path2): StochasticDepth(p=0.0, mode=row)
              (norm2): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
            )
          )
        )
        (pooler): XLMRobertaPooler(
          (dense): ParametrizedLinear(
            in_features=1024, out_features=1024, bias=True
            (parametrizations): ModuleDict(
              (weight): ParametrizationList(
                (0): LoRAParametrization()
              )
            )
          )
          (activation): Tanh()
        )
      )
    )
  )
  (pooler): Pooling({'word_embedding_dimension': 1024, '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})
  (normalizer): 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("Jrinky/final_stage1")
# Run inference
sentences = [
    'What items are present in the described setting along with the policy on pets',
    'There is also a grandfather clock and two oriental lions on grey marbled pedestals. Pets are allowed (Charges may be applicable)',
    'Investigators stated that Philoumenos appeared to have been trying to protect his face with his hands when a blow to his face or head severed one finger on each hand. Raby escaped the scene of the crime undetected. Raby was subsequently found to have acted alone, "without any connection to a religious or political entity." An investigation launched by the Israeli police initially failed to identify the killer. Raby was arrested on 17 November 1982 as he again attempted enter the Monastery at Jacob\'s Well illicitly by climbing over a wall; he was carrying hand grenades. Raby supplied the police with accurate details of his earlier, previously unsolved, crimes. These were the murder of Philoumenos; a March 1979 murder of a Jewish gynecologist in Tel-Aviv; the murder of the family of a woman in Lod, Israel in April 1979 who claimed to have clairvoyant powers; and an assault on a nun at the Jacob\'s Well holy site in April 1982. The nun was seriously wounded in the attack. Both she and the gynecologist were attacked by axe, according to prosecutors. Raby, a newly religious Jew, was described as unwashed, dressed in worn-out clothing, and audibly muttered passages of scripture in a strange manner. Psychiatric evaluations found that he was mentally incompetent to stand trial; he was committed to a mental hospital; details of his subsequent whereabouts are restricted by privacy regulations. At a court hearing after his arrest, an Israeli prosecutor told the court that Raby was convinced that the monastery was the site of the ancient Jewish Temple, and that he made an attempt on the life of the nun "in response to a divine command." Erroneous accounts\nInitial accounts depicted the murder as an anti-Christian hate attack carried out by a group of Jewish settlers, the result being what Maariv described as "a wave of hatred" in Greece. Reports indicating that "radical Jews" had tortured Philoumenos and "cut off the fingers of his hand" before killing him had appeared in Greek newspapers. Maariv also quoted an official in the Greek Orthodox Patriarchate in Jerusalem asserting that "the murder was carried out by radical religious Jews" claiming that "the Well does not belong to Christians but to Jews". In a 2017 article in the journal Israel Studies, researchers David Gurevich and Yisca Harani found that false accounts blaming the slaying on "settlers" and "Zionist extremists" persisted even after the arrest of the assailant and his confinement in a mental institution, and that there were "patterns of ritual murder accusation in the popular narrative." The same theme was echoed in parts of the Eastern Orthodox community and by some secular sources, including Blackwell\'s Dictionary of Eastern Christianity, the Encyclopedia of the Israeli-Palestinian Conflict,  The Spectator and Times Literary Supplement, as well as Wikipedia. Gurevich and Harani contended that a 1989 account of the murder, published in Orthodox America, a publication of the Russian Orthodox Church Outside Russia, became the basis of an anti-Semitic ritual murder narrative, according to which a group of anti-Christianity Jews first harassed Philoumenos and destroyed Christian holy objects at the monastery, then murdered him. Veneration\nIn 2009 the Greek Orthodox Patriarchate of Jerusalem recognised him as a holy martyr of the Eastern Orthodox Church, thirty years after his "martyrdom". The "careful" wording of the pronouncement of the Jerusalem Patriarchate that canonized Philoumenos makes no mention of murderer\'s faith or ethnicity; he is described as a "vile man" a "heterodox fanatic visitor" and, inaccurately, as an individual who "with an axe, opened a deep cut across his forehead, cut off the fingers of his right hand, and upon escaping threw a grenade which ended the Father\'s life."',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Training Details

Training Dataset

Unnamed Dataset

Size: 199,321 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 7 tokens
mean: 17.09 tokens
max: 46 tokens
min: 8 tokens
mean: 109.45 tokens
max: 1835 tokens

Samples:

anchor	positive
`Where is Nagpada located`	`Nagpada is a neighbourhood in South Mumbai.`
`What types of players are associated with Folkestone F.C., Midland Football League, and English Football League`	`players[object Object]Folkestone F.C. players[object Object]Midland Football League players[object Object]English Football League players`
`What is Anthony Elujoba known for in the field of Pharmacognosy`	`Anthony Elujoba (born 1948) is a Nigerian professor of Pharmacognosy, fondly referred to as the "village chemist" because of his involvement in research into medicinal plants. He was acting vice chancellor of Obafemi Awolowo University, Nigeria.`

Loss: cachedselfloss.CachedInfonce with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Evaluation Dataset

Unnamed Dataset

Size: 4,068 evaluation samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 6 tokens
mean: 17.26 tokens
max: 40 tokens
min: 6 tokens
mean: 108.28 tokens
max: 2233 tokens

Samples:

anchor	positive
`What metaphor is being used to describe collaboration in the text segment`	`Like two oxen in a field, tied shoulder to shoulder. With Jesus doing all of the heavy lifting.`
`What titles did McGurk win while playing as a schoolboy and a student`	`He won consecutive MacRory Cup titles lining out as a schoolboy with St Patrick's College, Maghera before winning a Sigerson Cup title as a student at Queen's University Belfast. McGurk progressed onto the Lavey senior teams in both codes and was corner-forward on the team that won the All-Ireland SCFC title in 1991.`
`What are the borders of the Trinity-Bellwoods neighborhood in Toronto`	`Trinity-Bellwoods is an inner city neighbourhood in Toronto, Ontario, Canada. It is bounded on the east by Bathurst Street, on the north by College Street, on the south by Queen Street West, and by Dovercourt Road on the west.`

Loss: cachedselfloss.CachedInfonce with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 2000
per_device_eval_batch_size: 2000
learning_rate: 2e-05
num_train_epochs: 1
warmup_ratio: 0.1
bf16: True
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: 2000
per_device_eval_batch_size: 2000
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: 1
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: 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: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
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
average_tokens_across_devices: False
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.1	10	0.3939	0.4079
0.2	20	0.4225	0.3920
0.3	30	0.4067	0.3819
0.4	40	0.3918	0.3760
0.5	50	0.4631	0.3719
0.6	60	0.3806	0.3686
0.7	70	0.3971	0.3663
0.8	80	0.3788	0.3655
0.9	90	0.3852	0.3649
1.0	100	0.3881	0.3648

Framework Versions

Python: 3.11.8
Sentence Transformers: 3.4.1
Transformers: 4.49.0
PyTorch: 2.4.0+cu121
Accelerate: 1.4.0
Datasets: 3.3.2
Tokenizers: 0.21.0

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",
}

CachedInfonce

bibtex

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

	anchor	positive
type	string	string
details	min: 7 tokens mean: 17.09 tokens max: 46 tokens	min: 8 tokens mean: 109.45 tokens max: 1835 tokens