rossieRuby/nyayadrishti-bert-v2

SentenceTransformer

This is a sentence-transformers model trained. 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
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 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(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (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})
)

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("rossieRuby/nyayadrishti-bert-v2")
# Run inference
sentences = [
    'What entry is mentioned in relation to tax on land?',
    'In that connection, the following observations lay down the guide lines: "It is well settled that the entries in the three legislative lists have to be interpreted in their widest amplitude and there fore if a tax can reasonably be held to be a tax on land it will come within entry 49. Further it is equally well settled that tax on land may be based on the annual value of the land and would still be a tax on land and would not be beyond the com petence of the State legislature on the ground that it is a tax on income: (See Ralla Ram vs The province of East Punjab: It follows therefore that the use to which the land is put can be taken into account in imposing a tax (1) ; (2)Quoted in Liberty cinema: P. 484. 339 on it within the meaning of entry 49 of List II, for the annual value of land which can certainly be taken into account in imposing a tax for the purpose of this entry would necessarily depend upon the use to which the land is put." (p. 49).',
    'Now it is true that the so called will of 1864 does not make provision for the grandsons, nor does it expressly confer an absolute estate on the legatees, but the witness is illiterate and had to depend on what he was told about the contents and meaning of the document, and what we have to test is the truth of his assertion that the plaintiff Mukand Ram and Kanhaiya Lal, and other members of the family, told him that Mst. Pato had given the property to her daughters and grandsons. If they told him this, as he says they did, then it operates as an admission against Mukand Ram and shifts the burden of proof to him because he was one of the persons who made the statement. The statements made by the others are not relevant except in so far as they prove the conduct of the family. The plaintiff (P.W. 11 in C.A. 91/50) admits that Mst. Pato divided the estate but says that it was only for convenience of management and that neither she nor her daughters had, or pretended to have, anything more than a life estate. He denies that there was any gift or family arrangement. But he had to admit that the grandsons also got properties at the same time. His explanation is that it was for the purposes of "shradh" and pilgrimage to Gaya and he says that though they were given possession they were not the "owners".',
]
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]

Training Details

Training Dataset

Unnamed Dataset

• Size: 29,992 training samples
• Columns: sentence_0, sentence_1, and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 7 tokens mean: 15.09 tokens max: 45 tokens	min: 10 tokens mean: 153.59 tokens max: 512 tokens	min: 0.0 mean: 0.48 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  Why did the respondents dispute the demurrage charges?	The respondents disputed the right of the Port Trust to charge any demurrage for the period during which the goods were detained by the Customs authorities for analytical test. as well as for the Import Trade Control formalities.	1.0
  What was the paid up capital of the subsidiary Company?	The paid up capital of the subsidiary Company was Rs. 7,91,100 divided into 7,911 shares of Rs. 100 each.	1.0
  How many parts does the schedule attached to the Act have?	The schedule attached to the Act specifies, under two parts, the employments in respect of which the minimum wages of the employees can be fixed; and section 27 authorises the "appropriate Government", after giving three months ' notice of its intention to do so, to add to either part of the schedule, any other employment, in respect of which it is of the opinion that minimum rates of wages should be fixed under the Act.	1.0

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 1
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• 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: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• num_train_epochs: 1
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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}
• tp_size: 0
• 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
• 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: batch_sampler
• multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss
0.2667	500	0.2532
0.5333	1000	0.2527
0.8	1500	0.2524

Framework Versions

• Python: 3.11.12
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.6.0
• Datasets: 2.14.4
• Tokenizers: 0.21.1

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",
}