elsayovita/bge-micro-v2-esg

BGE micro v2 ESG

This is a sentence-transformers model finetuned from TaylorAI/bge-micro-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: TaylorAI/bge-micro-v2
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 tokens
• Similarity Function: Cosine Similarity
• 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': 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})
)

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("elsayovita/bge-micro-v2-esg")
# Run inference
sentences = [
    'Employee health and well-being has never been more topical than it was in the past year. We understand that people around the world, including our employees, have been increasingly exposed to factors affecting their physical and mental wellbeing. We are committed to creating an environment that supports our employees and ensures they feel valued and have a sense of belonging. We utilised',
    "Question: What is the company's commitment towards its employees' health and well-being based on the provided context information?",
    'What types of skills does NetLink focus on developing through their training and development opportunities for employees?',
]
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]

Evaluation

Metrics

Information Retrieval

• Dataset: dim_384
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7394
cosine_accuracy@3	0.8871
cosine_accuracy@5	0.9144
cosine_accuracy@10	0.9383
cosine_precision@1	0.7394
cosine_precision@3	0.2957
cosine_precision@5	0.1829
cosine_precision@10	0.0938
cosine_recall@1	0.0205
cosine_recall@3	0.0246
cosine_recall@5	0.0254
cosine_recall@10	0.0261
cosine_ndcg@10	0.1866
cosine_mrr@10	0.8176
cosine_map@100	0.0228

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7316
cosine_accuracy@3	0.8832
cosine_accuracy@5	0.9112
cosine_accuracy@10	0.9355
cosine_precision@1	0.7316
cosine_precision@3	0.2944
cosine_precision@5	0.1822
cosine_precision@10	0.0936
cosine_recall@1	0.0203
cosine_recall@3	0.0245
cosine_recall@5	0.0253
cosine_recall@10	0.026
cosine_ndcg@10	0.1855
cosine_mrr@10	0.812
cosine_map@100	0.0226

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7171
cosine_accuracy@3	0.8736
cosine_accuracy@5	0.9013
cosine_accuracy@10	0.9279
cosine_precision@1	0.7171
cosine_precision@3	0.2912
cosine_precision@5	0.1803
cosine_precision@10	0.0928
cosine_recall@1	0.0199
cosine_recall@3	0.0243
cosine_recall@5	0.025
cosine_recall@10	0.0258
cosine_ndcg@10	0.183
cosine_mrr@10	0.7997
cosine_map@100	0.0223

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6759
cosine_accuracy@3	0.836
cosine_accuracy@5	0.8714
cosine_accuracy@10	0.9061
cosine_precision@1	0.6759
cosine_precision@3	0.2787
cosine_precision@5	0.1743
cosine_precision@10	0.0906
cosine_recall@1	0.0188
cosine_recall@3	0.0232
cosine_recall@5	0.0242
cosine_recall@10	0.0252
cosine_ndcg@10	0.1755
cosine_mrr@10	0.7621
cosine_map@100	0.0212

Information Retrieval

• Dataset: dim_32
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5759
cosine_accuracy@3	0.7347
cosine_accuracy@5	0.7802
cosine_accuracy@10	0.8298
cosine_precision@1	0.5759
cosine_precision@3	0.2449
cosine_precision@5	0.156
cosine_precision@10	0.083
cosine_recall@1	0.016
cosine_recall@3	0.0204
cosine_recall@5	0.0217
cosine_recall@10	0.0231
cosine_ndcg@10	0.1552
cosine_mrr@10	0.6648
cosine_map@100	0.0186

Training Details

Training Dataset

Unnamed Dataset

• Size: 11,863 training samples
• Columns: context and question
• Approximate statistics based on the first 1000 samples:

  	context	question
  type	string	string
  details	min: 13 tokens mean: 40.74 tokens max: 277 tokens	min: 11 tokens mean: 24.4 tokens max: 62 tokens

• Samples:

  context	question
  The engagement with key stakeholders involves various topics and methods throughout the year	Question: What does the engagement with key stakeholders involve throughout the year?
  For unitholders and analysts, the focus is on business and operations, the release of financial results, and the overall performance and announcements	Question: What is the focus for unitholders and analysts in terms of business and operations, financial results, performance, and announcements?
  These are communicated through press releases and other required disclosures via SGXNet and NetLink's website	What platform is used to communicate press releases and required disclosures for NetLink?

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          384,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 2
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: False
• load_best_model_at_end: True
• optim: adamw_torch_fused
• 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: 32
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• eval_accumulation_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: 2
• 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: False
• 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
• 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: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_128_cosine_map@100	dim_256_cosine_map@100	dim_32_cosine_map@100	dim_384_cosine_map@100	dim_64_cosine_map@100
0.4313	10	5.0772	-	-	-	-	-
0.8625	20	3.2666	-	-	-	-	-
1.0350	24	-	0.0221	0.0224	0.0185	0.0226	0.0211
1.2264	30	3.1157	-	-	-	-	-
1.6577	40	2.585	-	-	-	-	-
1.9164	46	-	0.0223	0.0226	0.0186	0.0228	0.0212

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.42.4
• PyTorch: 2.4.0+cu121
• Accelerate: 0.32.1
• Datasets: 2.21.0
• 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",
}

MatryoshkaLoss

@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

@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}
}