You can deploy this model and start earning money today!

SentenceTransformer based on BAAI/bge-base-en-v1.5

Name: votum-acts-v1
Availability: InStock

This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5 on the json 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: BAAI/bge-base-en-v1.5
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- json
Language: en
License: apache-2.0

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': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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("Tejasw1/votum-acts-v1")
# Run inference
sentences = [
    'Represent this sentence for searching relevant passages: According to **Section 13(2)(a) of the Central Goods and Services Tax Act, 2017**, what is the time of supply of services if the invoice is issued within the prescribed period?',
    'Document: (1) The liability to pay tax on services shall arise at the time ofsupply, as determined in accordance with the provisions of this section.(2) The time of supply of services shall be the earliest of the following dates, namely:--(a) the date of issue of invoice by the supplier, if the invoice is issued within the periodprescribed under 1*** sub-section (2) of section 31 or the date of receipt of payment, whichever is earlier; or(b) the date of provision of service, if the invoice is not issued within the period prescribed under sub-section (2) of 1*** section 31 or the date of receipt of payment, whichever is earlier; or(c) the date on which the recipient shows the receipt of services in his books of account, in a case where the provisions of clause (a) or clause (b) do not apply:Provided that where the supplier of taxable service receives an amount up to one thousand rupees inexcess of the amount indicated in the tax invoice, the time of supply to the extent of such excess amountshall, at the option of the said supplier, be the date of issue of invoice relating to such excess amount.Explanation.--For the purposes of clauses (a) and (b)--(i) the supply shall be deemed to have been made to the extent it is covered by the invoice or, asthe case may be, the payment;(ii) "the date of receipt of payment" shall be the date on which the payment is entered in thebooks of account of the supplier or the date on which the payment is credited to his bank account,whichever is earlier.(3) In case of supplies in respect of which tax is paid or liable to be paid on reverse charge basis, thetime of supply shall be the earlier of the following dates, namely:--(a) the date of payment as entered in the books of account of the recipient or the date on which thepayment is debited in his bank account, whichever is earlier; or(b) the date immediately following sixty days from the date of issue of invoice or any otherdocument, by whatever name called, in lieu thereof by the supplier:Provided that where it is not possible to determine the time of supply under clause (a) or clause (b),the time of supply shall be the date of entry in the books of account of the recipient of supply:Provided further that in case of supply by associated enterprises, where the supplier of service islocated outside India, the time of supply shall be the date of entry in the books of account of the recipientof supply or the date of payment, whichever is earlier.(4) In case of supply of vouchers by a supplier, the time of supply shall be--(a) the date of issue of voucher, if the supply is identifiable at that point; or(b) the date of redemption of voucher, in all other cases.(5) Where it is not possible to determine the time of supply under the provisions of sub-section (2) orsub-section (3) or sub-section (4), the time of supply shall--(a) in a case where a periodical return has to be filed, be the date on which such return is to befiled; or(b) in any other case, be the date on which the tax is paid.(6) The time of supply to the extent it relates to an addition in the value of supply by way of interest,late fee or penalty for delayed payment of any consideration shall be the date on which the supplierreceives such addition in value.\t\t\t\t\t\t\t\t\t1. The words, brackets and figure “sub-section (2) of” omitted by Act 31 of 2018, s. 7 (w.e.f. 1-2-2019).',
    'Document: Article (51) Prohibitions                                                                                                        Public Welfare Associations and their Members may not do the following:\r1. Practice any Public Welfare activity other than those stipulated in its By-laws.\r2. Practice any political or partisan activity, collecting information, interfering in politics or matters affecting the security of the State and its law of government, or using its Office for that purpose, or provoking sectarian, racial, or religious disputes.\r3. Affiliate, join, participate in, or deal with any illegal Associations or entities, or any natural or Legal Person belonging to it, whether inside or outside the State, or financing or providing support to them in any way.\r4. Deal with, financing, or providing support to any illegal Association, terrorist Association, or entity, or any natural or Legal Person belonging to any of them.\r5. Form secret societies, companies, or formations of a secret, military, or paramilitary nature, or calling for favouring, supporting, or financing violence or terrorist organisations.\r6. Practice activities that would disturb public order, public morals, Emirati customs and traditions, or threaten the national security of the State.\r7. Call for discrimination between citizens or residents of the State on the basis of gender, origin, colour, language, religion or belief, or any activity that calls for racism, incitement to hatred, or other reasons that are contrary to the Constitution and the legislation in force in the State.\r8. Participate in supporting or financing the electoral campaigns of any candidate in elections and referendums, or presenting a candidate in those elections on behalf of the Association.\r9. Grant any professional or applied certificates without authorisation from the Competent Authorities in the State, or without an official partnership with one of the specialised universities or the Competent Authorities, and in accordance with the rules regulating this in the State.\r10. Practice any Public Welfare Activities outside the spatial scope of the licence issued to him by the Competent Authority.\r11. Practice any Activities that require a licence or approval from a governmental entity, before obtaining a licence or approval from that entity and the Competent Authority.\r12. Aim to make a profit for the Members of a Public Welfare Association, or engaging in an activity aimed at that, or distributing the Funds of a Public Welfare Association to its Members, employees, or those responsible for its management.\r13. Conduct opinion polls, publishing or making their results available, or conducting field research or presenting their results, without obtaining prior approval from the Ministry and the relevant authorities in the State.\r14. Conclude agreement in any form with a foreign party outside the State before the Ministry approval, as well as any amendment to it.\r15. Deal in any way with embassies, consulates and diplomatic missions without obtaining permission from the Competent Authority, and without the approval of the Ministry of Foreign Affairs in accordance with the procedures followed in this regard.\r16. Open branches or Offices outside the State.\r17. Interfere in the work of any State or Local Government Authority.\r18. Represent any individual or group before the Court in any lawsuits related to the interests of these individuals or groups.\r19. Raise and disseminate information that urges non-respect for the Constitution, laws and legislation in force in the State, non-respect for judicial rulings, or prevention of their implementation.\r20. Publish information, news, or propaganda that would prejudice public order or harm the public interest, public security, or public morals.\r21. Hold courses, workshops, Meetings or seminars, whether inside or outside the State, that would harm public order, harm the public interest or public security, or harm public morals.\r22. Work in any way under political cover.\r23. Any other prohibitions in implementation of the legislation in force in the State.\r24. Any other prohibitions determined by the Competent Authority, pursuant to the Resolutions issued by it in this regard.',
]
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

Information Retrieval

Datasets: dim_768 and dim_512
Evaluated with InformationRetrievalEvaluator

Metric	dim_768	dim_512
cosine_accuracy@1	0.0178	0.0153
cosine_accuracy@3	0.2606	0.2606
cosine_accuracy@5	0.4958	0.4941
cosine_accuracy@10	0.7156	0.7114
cosine_precision@1	0.0178	0.0153
cosine_precision@3	0.0869	0.0869
cosine_precision@5	0.0992	0.0988
cosine_precision@10	0.0716	0.0711
cosine_recall@1	0.0178	0.0153
cosine_recall@3	0.2606	0.2606
cosine_recall@5	0.4958	0.4941
cosine_recall@10	0.7156	0.7114
cosine_ndcg@10	0.3202	0.3172
cosine_mrr@10	0.1982	0.1957
cosine_map@100	0.2108	0.2082

Training Details

Training Dataset

json

Dataset: json
Size: 22,370 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 22 tokens
mean: 41.92 tokens
max: 77 tokens
min: 4 tokens
mean: 264.59 tokens
max: 512 tokens

Samples:

anchor	positive
`Represent this sentence for searching relevant passages: Under Section 52(1)(d) of the Bharatiya Sakshya Adhiniyam, 2023, are courts required to take judicial notice of the seals of all courts and tribunals?`	Document: (1) The Court shall take judicial notice of thefollowing facts, namely:--(a) all laws in force in the territory of India including laws having extra-territorial operation;(b) international treaty, agreement or convention with country or countries by India, or decisionsmade by India at international associations or other bodies;(c) the course of proceeding of the Constituent Assembly of India, of Parliament of India and ofthe State Legislatures;(d) the seals of all Courts and Tribunals;(e) the seals of Courts of Admiralty and Maritime Jurisdiction, Notaries Public, and all sealswhich any person is authorised to use by the Constitution, or by an Act of Parliament or StateLegislatures, or Regulations having the force of law in India;(f) the accession to office, names, titles, functions, and signatures of the persons filling for thetime being any public office in any State, if the fact of their appointment to such office is notified inany Official Gazette;(g) the existence, title...
`Represent this sentence for searching relevant passages: Is it permissible for a bankruptcy trustee to appoint the bankrupt to supervise the management of the estate, carry on his business, or assist in administering the estate under the Insolvency and Bankruptcy Code, 2016, Section 153?`	Document: The bankruptcy trustee for the purposes of thisChapter may after procuring the approval of the committee of creditors,— (a) carry on any business of the bankrupt as far as may be necessary for winding it upbeneficially;(b) bring, institute or defend any legal action or proceedings relating to the property comprised inthe estate of the bankrupt;(c) accept as consideration for the sale of any property a sum of money due at a future timesubject to certain stipulations such as security;(d) mortgage or pledge any property for the purpose of raising money for the payment of thedebts of the bankrupt;(e) where any right, option or other power forms part of the estate of the bankrupt, makepayments or incur liabilities with a view to obtaining, for the benefit of the creditors, any propertywhich is the subject of such right, option or power;(f) refer to arbitration or compromise on such terms as may be agreed, any debts subsisting orsupposed to subsist between the bankrupt and any pers...
`Represent this sentence for searching relevant passages: What insurance requirements are imposed on Federal Agencies occupying Union Owned Properties under Article (23) of the Federal Decree Concerning the Union Owned Properties?`	`Document: Article (23) Obligations of the Federal Authorities that occupy any of the Union Owned Properties 1. In addition to the obligations stipulated herein, every Federal Agency that occupies, manages, or supervises the management of any of the Union Owned Properties shall comply, as follows:`
a. Provide a report showing the legal and surveying status of that property, estimating its value, and indicating its architectural and constructional condition, along with attaching its construction plan and any data or any facts, documents or papers related in any way to the sources of its ownership or occupancy, within a period not exceeding (6) six months from the effective date herein. His authority shall provide the Ministry with a copy of this report immediately upon completion of its preparation, and it shall renew this data and provide the Ministry with a copy of it whenever necess...

Loss: MatryoshkaLoss with these parameters:
json
1{ 2 "loss": "MultipleNegativesRankingLoss", 3 "matryoshka_dims": [ 4 768, 5 512 6 ], 7 "matryoshka_weights": [ 8 1, 9 1 10 ], 11 "n_dims_per_step": -1 12}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
gradient_accumulation_steps: 8
learning_rate: 2e-05
num_train_epochs: 4
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: True
tf32: True
load_best_model_at_end: True
optim: adamw_torch_fused
prompts: {'anchor': 'Represent this sentence for searching relevant passages: ', 'positive': 'Document: '}
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 8
per_device_eval_batch_size: 8
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 8
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: 4
max_steps: -1
lr_scheduler_type: cosine
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: True
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: True
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_fused
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
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: {'anchor': 'Represent this sentence for searching relevant passages: ', 'positive': 'Document: '}
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	dim_768_cosine_ndcg@10	dim_512_cosine_ndcg@10
0.0286	10	0.5889	-	-
0.0572	20	0.4858	-	-
0.0858	30	0.4432	-	-
0.1144	40	0.3437	-	-
0.1430	50	0.2103	-	-
0.1716	60	0.1903	-	-
0.2002	70	0.1414	-	-
0.2288	80	0.1627	-	-
0.2574	90	0.1609	-	-
0.2860	100	0.0968	-	-
0.3146	110	0.1367	-	-
0.3432	120	0.1228	-	-
0.3718	130	0.0891	-	-
0.4004	140	0.1116	-	-
0.4290	150	0.1173	-	-
0.4576	160	0.1162	-	-
0.4862	170	0.1124	-	-
0.5148	180	0.1014	-	-
0.5434	190	0.0767	-	-
0.5720	200	0.0745	-	-
0.6006	210	0.0691	-	-
0.6292	220	0.094	-	-
0.6578	230	0.0692	-	-
0.6864	240	0.0471	-	-
0.7151	250	0.0647	-	-
0.7437	260	0.077	-	-
0.7723	270	0.0551	-	-
0.8009	280	0.0538	-	-
0.8295	290	0.0863	-	-
0.8581	300	0.0698	-	-
0.8867	310	0.0599	-	-
0.9153	320	0.0494	-	-
0.9439	330	0.0746	-	-
0.9725	340	0.0544	-	-
0.9982	349	-	0.3143	0.3102
1.0021	350	0.06	-	-
1.0307	360	0.09	-	-
1.0593	370	0.0597	-	-
1.0880	380	0.0613	-	-
1.1166	390	0.0589	-	-
1.1452	400	0.0309	-	-
1.1738	410	0.0378	-	-
1.2024	420	0.0417	-	-
1.2310	430	0.0417	-	-
1.2596	440	0.0412	-	-
1.2882	450	0.0214	-	-
1.3168	460	0.0374	-	-
1.3454	470	0.0388	-	-
1.3740	480	0.0188	-	-
1.4026	490	0.0247	-	-
1.4312	500	0.0275	-	-
1.4598	510	0.0336	-	-
1.4884	520	0.017	-	-
1.5170	530	0.0234	-	-
1.5456	540	0.0163	-	-
1.5742	550	0.0193	-	-
1.6028	560	0.0209	-	-
1.6314	570	0.0252	-	-
1.6600	580	0.02	-	-
1.6886	590	0.0199	-	-
1.7172	600	0.0162	-	-
1.7458	610	0.0246	-	-
1.7744	620	0.0133	-	-
1.8030	630	0.017	-	-
1.8316	640	0.0241	-	-
1.8602	650	0.018	-	-
1.8888	660	0.0186	-	-
1.9174	670	0.0121	-	-
1.9460	680	0.0264	-	-
1.9746	690	0.0112	-	-
1.9975	698	-	0.3174	0.3161
2.0043	700	0.0159	-	-
2.0329	710	0.0295	-	-
2.0615	720	0.0197	-	-
2.0901	730	0.0252	-	-
2.1187	740	0.019	-	-
2.1473	750	0.0074	-	-
2.1759	760	0.0122	-	-
2.2045	770	0.0116	-	-
2.2331	780	0.0113	-	-
2.2617	790	0.0132	-	-
2.2903	800	0.0112	-	-
2.3189	810	0.0167	-	-
2.3475	820	0.0078	-	-
2.3761	830	0.0079	-	-
2.4047	840	0.0072	-	-
2.4333	850	0.008	-	-
2.4619	860	0.0135	-	-
2.4905	870	0.0087	-	-
2.5191	880	0.0066	-	-
2.5477	890	0.0052	-	-
2.5763	900	0.0077	-	-
2.6049	910	0.0084	-	-
2.6335	920	0.0096	-	-
2.6621	930	0.0067	-	-
2.6907	940	0.0072	-	-
2.7193	950	0.0061	-	-
2.7479	960	0.0132	-	-
2.7765	970	0.0061	-	-
2.8051	980	0.0058	-	-
2.8338	990	0.01	-	-
2.8624	1000	0.0084	-	-
2.8910	1010	0.0082	-	-
2.9196	1020	0.0055	-	-
2.9482	1030	0.0073	-	-
2.9768	1040	0.0074	-	-
2.9968	1047	-	0.323	0.3161
3.0064	1050	0.0086	-	-
3.0350	1060	0.0127	-	-
3.0636	1070	0.0083	-	-
3.0922	1080	0.0111	-	-
3.1208	1090	0.0091	-	-
3.1494	1100	0.0037	-	-
3.1780	1110	0.0074	-	-
3.2066	1120	0.005	-	-
3.2353	1130	0.006	-	-
3.2639	1140	0.0071	-	-
3.2925	1150	0.0062	-	-
3.3211	1160	0.008	-	-
3.3497	1170	0.0042	-	-
3.3783	1180	0.003	-	-
3.4069	1190	0.0049	-	-
3.4355	1200	0.004	-	-
3.4641	1210	0.0062	-	-
3.4927	1220	0.0056	-	-
3.5213	1230	0.0048	-	-
3.5499	1240	0.0034	-	-
3.5785	1250	0.0045	-	-
3.6071	1260	0.0041	-	-
3.6357	1270	0.0048	-	-
3.6643	1280	0.0045	-	-
3.6929	1290	0.0044	-	-
3.7215	1300	0.0047	-	-
3.7501	1310	0.0061	-	-
3.7787	1320	0.0037	-	-
3.8073	1330	0.0045	-	-
3.8359	1340	0.0068	-	-
3.8645	1350	0.0048	-	-
3.8931	1360	0.0056	-	-
3.9217	1370	0.0049	-	-
3.9503	1380	0.0055	-	-
3.9789	1390	0.004	-	-
3.9961	1396	-	0.3202	0.3172

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.11.5
Sentence Transformers: 3.3.1
Transformers: 4.46.3
PyTorch: 2.4.1+cu121
Accelerate: 0.34.2
Datasets: 3.0.0
Tokenizers: 0.20.3

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",
}

MatryoshkaLoss

bibtex

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

bibtex

@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}
}

	anchor	positive
type	string	string
details	min: 22 tokens mean: 41.92 tokens max: 77 tokens	min: 4 tokens mean: 264.59 tokens max: 512 tokens