KayaTechAI/BGE-M3-0.56B-Fine-Tuned-Telecom-Technical-Documents-Retrieval-Embedding-Generalization-Baseline

BGE-M3-Telecom-Retrieval-Embedding

This is a sentence-transformers model finetuned from BAAI/bge-m3 on the telecom-technical-documents-retrieval-embedding-dataset dataset. 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
• Base model: BAAI/bge-m3
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • telecom-technical-documents-retrieval-embedding-dataset
• 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': 8192, 'do_lower_case': False, 'architecture': 'XLMRobertaModel'})
  (1): Pooling({'word_embedding_dimension': 1024, '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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("KayaTechAI/BGE-M3-0.56B-Fine-Tuned-Telecom-Technical-Documents-Retrieval-Embedding-Generalization-Baseline")
# Run inference
sentences = [
    'What is the provisioning scope for the eMLPP service?',
    'eMLPP is provisioned per subscriber.',
    'The main objective is to verify that the User Equipment (UE) tracks channel variations and selects the optimal transport format for frequency non-selective scheduling.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000,  0.7420, -0.0991],
#         [ 0.7420,  1.0000, -0.1267],
#         [-0.0991, -0.1267,  1.0000]])

Evaluation

Metrics

Information Retrieval

• Dataset: dim_1024
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 1024
  }
  

Metric	Value
cosine_accuracy@1	0.7852
cosine_accuracy@3	0.9008
cosine_accuracy@5	0.9324
cosine_accuracy@10	0.956
cosine_precision@1	0.7852
cosine_precision@3	0.3003
cosine_precision@5	0.1865
cosine_precision@10	0.0956
cosine_recall@1	0.7852
cosine_recall@3	0.9008
cosine_recall@5	0.9324
cosine_recall@10	0.956
cosine_ndcg@10	0.8744
cosine_mrr@10	0.8478
cosine_map@100	0.8497

Information Retrieval

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 768
  }
  

Metric	Value
cosine_accuracy@1	0.7848
cosine_accuracy@3	0.9008
cosine_accuracy@5	0.9312
cosine_accuracy@10	0.9572
cosine_precision@1	0.7848
cosine_precision@3	0.3003
cosine_precision@5	0.1862
cosine_precision@10	0.0957
cosine_recall@1	0.7848
cosine_recall@3	0.9008
cosine_recall@5	0.9312
cosine_recall@10	0.9572
cosine_ndcg@10	0.8748
cosine_mrr@10	0.848
cosine_map@100	0.8499

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 512
  }
  

Metric	Value
cosine_accuracy@1	0.784
cosine_accuracy@3	0.9012
cosine_accuracy@5	0.9308
cosine_accuracy@10	0.9544
cosine_precision@1	0.784
cosine_precision@3	0.3004
cosine_precision@5	0.1862
cosine_precision@10	0.0954
cosine_recall@1	0.784
cosine_recall@3	0.9012
cosine_recall@5	0.9308
cosine_recall@10	0.9544
cosine_ndcg@10	0.8734
cosine_mrr@10	0.8469
cosine_map@100	0.8488

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 256
  }
  

Metric	Value
cosine_accuracy@1	0.7768
cosine_accuracy@3	0.8956
cosine_accuracy@5	0.9292
cosine_accuracy@10	0.9524
cosine_precision@1	0.7768
cosine_precision@3	0.2985
cosine_precision@5	0.1858
cosine_precision@10	0.0952
cosine_recall@1	0.7768
cosine_recall@3	0.8956
cosine_recall@5	0.9292
cosine_recall@10	0.9524
cosine_ndcg@10	0.8691
cosine_mrr@10	0.8418
cosine_map@100	0.8438

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 128
  }
  

Metric	Value
cosine_accuracy@1	0.7716
cosine_accuracy@3	0.888
cosine_accuracy@5	0.92
cosine_accuracy@10	0.9456
cosine_precision@1	0.7716
cosine_precision@3	0.296
cosine_precision@5	0.184
cosine_precision@10	0.0946
cosine_recall@1	0.7716
cosine_recall@3	0.888
cosine_recall@5	0.92
cosine_recall@10	0.9456
cosine_ndcg@10	0.8624
cosine_mrr@10	0.8353
cosine_map@100	0.8374

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 64
  }
  

Metric	Value
cosine_accuracy@1	0.7452
cosine_accuracy@3	0.86
cosine_accuracy@5	0.8992
cosine_accuracy@10	0.9332
cosine_precision@1	0.7452
cosine_precision@3	0.2867
cosine_precision@5	0.1798
cosine_precision@10	0.0933
cosine_recall@1	0.7452
cosine_recall@3	0.86
cosine_recall@5	0.8992
cosine_recall@10	0.9332
cosine_ndcg@10	0.8411
cosine_mrr@10	0.8114
cosine_map@100	0.8138

Training Details

Training Dataset

telecom-technical-documents-retrieval-embedding-dataset

• Dataset: telecom-technical-documents-retrieval-embedding-dataset at 3ebf34a
• Size: 127,731 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 8 tokens mean: 22.46 tokens max: 75 tokens	min: 5 tokens mean: 31.38 tokens max: 95 tokens

• Samples:

  anchor	positive
  What is the estimated Transmit power considered sufficient for achieving 95% Downlink coverage with a single Base Station?	Approximately 14 dBm Transmit power is considered sufficient.
  What is the primary goal of the Nominal Accuracy requirement?	The primary goal of the Nominal Accuracy requirement is to ensure good accuracy when signal conditions are ideal.
  What happens on the mobile station side if contention resolution fails because the G-RNTI value in the network's acknowledgement message differs from what the mobile station sent?	If the mobile station receives a PACKET UPLINK ACK/NACK message with a G-RNTI value different from the one it included in its first RLC data blocks, it signifies a contention resolution failure, and the mobile station will not transmit a PACKET CONTROL ACKNOWLEDGEMENT.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          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: 32
• gradient_accumulation_steps: 16
• 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
• 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: 32
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• 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: None
• hub_always_push: False
• hub_revision: None
• 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
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	dim_1024_cosine_ndcg@10	dim_768_cosine_ndcg@10	dim_512_cosine_ndcg@10	dim_256_cosine_ndcg@10	dim_128_cosine_ndcg@10	dim_64_cosine_ndcg@10
0.0401	10	1.8502	-	-	-	-	-	-
0.0802	20	1.54	-	-	-	-	-	-
0.1202	30	1.0721	-	-	-	-	-	-
0.1603	40	0.8011	-	-	-	-	-	-
0.2004	50	0.6764	-	-	-	-	-	-
0.2405	60	0.5557	-	-	-	-	-	-
0.2806	70	0.4677	-	-	-	-	-	-
0.3206	80	0.4346	-	-	-	-	-	-
0.3607	90	0.4035	-	-	-	-	-	-
0.4008	100	0.3237	-	-	-	-	-	-
0.4409	110	0.3119	-	-	-	-	-	-
0.4810	120	0.2999	-	-	-	-	-	-
0.5210	130	0.287	-	-	-	-	-	-
0.5611	140	0.2993	-	-	-	-	-	-
0.6012	150	0.3116	-	-	-	-	-	-
0.6413	160	0.2408	-	-	-	-	-	-
0.6814	170	0.2748	-	-	-	-	-	-
0.7214	180	0.2801	-	-	-	-	-	-
0.7615	190	0.2534	-	-	-	-	-	-
0.8016	200	0.2597	-	-	-	-	-	-
0.8417	210	0.2602	-	-	-	-	-	-
0.8818	220	0.2298	-	-	-	-	-	-
0.9218	230	0.2385	-	-	-	-	-	-
0.9619	240	0.2156	-	-	-	-	-	-
1.0	250	0.2006	0.8670	0.8656	0.8628	0.8567	0.8440	0.8126
1.0401	260	0.1423	-	-	-	-	-	-
1.0802	270	0.161	-	-	-	-	-	-
1.1202	280	0.1638	-	-	-	-	-	-
1.1603	290	0.1358	-	-	-	-	-	-
1.2004	300	0.1609	-	-	-	-	-	-
1.2405	310	0.1643	-	-	-	-	-	-
1.2806	320	0.1413	-	-	-	-	-	-
1.3206	330	0.1411	-	-	-	-	-	-
1.3607	340	0.1416	-	-	-	-	-	-
1.4008	350	0.1315	-	-	-	-	-	-
1.4409	360	0.1377	-	-	-	-	-	-
1.4810	370	0.1174	-	-	-	-	-	-
1.5210	380	0.1319	-	-	-	-	-	-
1.5611	390	0.1169	-	-	-	-	-	-
1.6012	400	0.115	-	-	-	-	-	-
1.6413	410	0.1382	-	-	-	-	-	-
1.6814	420	0.1576	-	-	-	-	-	-
1.7214	430	0.1253	-	-	-	-	-	-
1.7615	440	0.133	-	-	-	-	-	-
1.8016	450	0.1231	-	-	-	-	-	-
1.8417	460	0.1364	-	-	-	-	-	-
1.8818	470	0.1047	-	-	-	-	-	-
1.9218	480	0.1268	-	-	-	-	-	-
1.9619	490	0.1284	-	-	-	-	-	-
2.0	500	0.1134	0.8728	0.8718	0.8706	0.8660	0.8563	0.8323
2.0401	510	0.0787	-	-	-	-	-	-
2.0802	520	0.0765	-	-	-	-	-	-
2.1202	530	0.079	-	-	-	-	-	-
2.1603	540	0.0764	-	-	-	-	-	-
2.2004	550	0.0913	-	-	-	-	-	-
2.2405	560	0.0797	-	-	-	-	-	-
2.2806	570	0.0865	-	-	-	-	-	-
2.3206	580	0.0843	-	-	-	-	-	-
2.3607	590	0.0841	-	-	-	-	-	-
2.4008	600	0.0842	-	-	-	-	-	-
2.4409	610	0.0827	-	-	-	-	-	-
2.4810	620	0.0968	-	-	-	-	-	-
2.5210	630	0.0781	-	-	-	-	-	-
2.5611	640	0.0745	-	-	-	-	-	-
2.6012	650	0.0744	-	-	-	-	-	-
2.6413	660	0.0854	-	-	-	-	-	-
2.6814	670	0.0807	-	-	-	-	-	-
2.7214	680	0.0678	-	-	-	-	-	-
2.7615	690	0.0795	-	-	-	-	-	-
2.8016	700	0.0845	-	-	-	-	-	-
2.8417	710	0.0846	-	-	-	-	-	-
2.8818	720	0.0957	-	-	-	-	-	-
2.9218	730	0.0723	-	-	-	-	-	-
2.9619	740	0.0676	-	-	-	-	-	-
3.0	750	0.0804	0.8748	0.8736	0.8728	0.8703	0.8633	0.8394
3.0401	760	0.053	-	-	-	-	-	-
3.0802	770	0.0622	-	-	-	-	-	-
3.1202	780	0.0777	-	-	-	-	-	-
3.1603	790	0.07	-	-	-	-	-	-
3.2004	800	0.0662	-	-	-	-	-	-
3.2405	810	0.0704	-	-	-	-	-	-
3.2806	820	0.0722	-	-	-	-	-	-
3.3206	830	0.0663	-	-	-	-	-	-
3.3607	840	0.0631	-	-	-	-	-	-
3.4008	850	0.0616	-	-	-	-	-	-
3.4409	860	0.0639	-	-	-	-	-	-
3.4810	870	0.0595	-	-	-	-	-	-
3.5210	880	0.071	-	-	-	-	-	-
3.5611	890	0.0748	-	-	-	-	-	-
3.6012	900	0.0648	-	-	-	-	-	-
3.6413	910	0.067	-	-	-	-	-	-
3.6814	920	0.0625	-	-	-	-	-	-
3.7214	930	0.0619	-	-	-	-	-	-
3.7615	940	0.0612	-	-	-	-	-	-
3.8016	950	0.067	-	-	-	-	-	-
3.8417	960	0.0597	-	-	-	-	-	-
3.8818	970	0.0593	-	-	-	-	-	-
3.9218	980	0.0742	-	-	-	-	-	-
3.9619	990	0.0691	-	-	-	-	-	-
4.0	1000	0.067	0.8744	0.8748	0.8734	0.8691	0.8624	0.8411

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.2.3
• Transformers: 4.55.4
• PyTorch: 2.10.0+cu128
• Accelerate: 1.13.0
• Datasets: 3.6.0
• Tokenizers: 0.21.4

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}
}