telecom_bge_embedding_model

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

This is a sentence-transformers model finetuned from BAAI/bge-large-en-v1.5. 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-large-en-v1.5 <!-- at revision d4aa6901d3a41ba39fb536a557fa166f842b0e09 -->
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity

<!-- - **Training Dataset:** Unknown --> <!-- - **Language:** Unknown --> <!-- - **License:** Unknown -->

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': 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("ml-enthusiast13/telecom_bge_embedding_model")
# Run inference
sentences = [
    'In EVS (Enhanced Voice Services) codec, as per 3GPP standards, why is interpolation for obtaining LP parameters for each subframe performed on LSP (Line Spectral Pairs) parameters instead of directly on LP (Linear Prediction) parameters?',
    "['In the context of the EVS codec according to 3GPP specifications, why is interpolation for LP parameter retrieval in each subframe conducted on LSP parameters rather than directly on LP parameters?', 'Why does the EVS codec perform interpolation on LSP parameters instead of LP parameters when obtaining LP parameters for each subframe as per 3GPP standards?', 'For obtaining LP parameters in every subframe, why is interpolation executed on LSP parameters within the EVS codec framework defined by 3GPP?']",
    "['In the context of the EVS codec according to 3GPP specifications, why is interpolation for LP parameter retrieval in each subframe conducted directly on LP parameters instead of LSP parameters?', 'Why does the EVS codec perform interpolation on LP parameters rather than LSP parameters when obtaining LP parameters for each subframe as per 3GPP standards?', 'For obtaining LP parameters in every subframe, why is interpolation executed on both LP and LSP parameters simultaneously within the EVS codec framework defined by 3GPP?']",
]
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]

<!-- ### Direct Usage (Transformers) <details><summary>Click to see the direct usage in Transformers</summary> </details> --> <!-- ### Downstream Usage (Sentence Transformers) You can finetune this model on your own dataset. <details><summary>Click to expand</summary> </details> --> <!-- ### Out-of-Scope Use *List how the model may foreseeably be misused and address what users ought not to do with the model.* -->

Evaluation

Metrics

Triplet

• Dataset: use_triplet_all-nli-test
• Evaluated with <code>TripletEvaluator</code>

<!-- ## Bias, Risks and Limitations *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* --> <!-- ### Recommendations *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* -->

Training Details

Training Dataset

Unnamed Dataset

• Size: 9,623 training samples
• Columns: <code>query</code>, <code>positive</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	query	positive	negative
  type	string	string	string
  details	<ul><li>min: 13 tokens</li><li>mean: 38.09 tokens</li><li>max: 84 tokens</li></ul>	<ul><li>min: 53 tokens</li><li>mean: 132.93 tokens</li><li>max: 313 tokens</li></ul>	<ul><li>min: 58 tokens</li><li>mean: 114.72 tokens</li><li>max: 244 tokens</li></ul>

• Samples:

  query	positive	negative
  <code>What action does the MME take if the bearer resource allocation requested by the UE is accepted by the network?</code>	<code>["Upon acceptance of the UE's requested bearer resource allocation by the network, the MME initiates either a dedicated EPS bearer context activation procedure or an EPS bearer context modification procedure.", "When the network approves the UE's request for bearer resources, the MME proceeds to start either a dedicated EPS bearer context activation process or modify an existing EPS bearer context.", 'If the requested bearer resource allocation by the UE is granted by the network, the MME begins either a dedicated EPS bearer context activation procedure or modifies an existing EPS bearer context.']</code>	<code>["When the network accepts the UE's bearer resource allocation request, the MME sends a BEARER RESOURCE ALLOCATION REJECT message to the UE.", 'If the network approves the requested bearer resources from the UE, the MME stops timer T3480 and transitions to the PROCEDURE TRANSACTION INACTIVE state.', 'Upon acceptance of the bearer resource allocation by the network, the MME releases the PTI value assigned for this procedure.']</code>
  <code>What is the primary purpose of conducting a single point power reference measurement at the beginning and end of the UE's OTA antenna performance testing procedure?</code>	<code>['The main objective of performing a single point power reference measurement before and after testing is to track and adjust for any Tx-power drift in the DUT, which helps maintain precise sensitivity and power assessment.', 'Conducting single point power reference measurements at both the commencement and conclusion of OTA antenna performance tests serves to monitor and correct for Tx-power variations in the DUT, ensuring accurate measurement results.', 'The purpose of executing a single point power reference measurement at the start and finish of the testing procedure is to observe and adjust for any changes in the Tx-power of the DUT, thus maintaining the integrity of sensitivity and power readings.']</code>	<code>['The primary aim of conducting a single point power reference measurement before and after testing is to ensure that the DUT adheres to its battery usage specifications during transmission tests.', "Single point power reference measurements at both ends of the testing process are intended to confirm the DUT's compliance with maximum transmit power limits, rather than focusing on Tx-power drift.", 'The main reason for conducting a single point power reference measurement before and after testing is to calibrate the testing equipment across all frequency bands used during the procedure.']</code>
  <code>What is the primary purpose of the Nnef_AnalyticsExposure_Unsubscribe service operation in the 5G System (5GS)?</code>	<code>['The main function of the Nnef_AnalyticsExposure_Unsubscribe service operation in the 5GS is to cancel an existing subscription for analytics data.', 'In the context of 5G System, the Nnef_AnalyticsExposure_Unsubscribe operation allows a network function consumer to terminate its current subscription to analytics information using the Subscription Correlation ID.', 'The purpose of the Nnef_AnalyticsExposure_Unsubscribe service in 5GS is to enable the NF consumer to withdraw from an existing analytics information subscription by providing the necessary Subscription Correlation ID.']</code>	<code>['The Nnef_AnalyticsExposure_Unsubscribe service operation in 5G System is used to initiate a new subscription for receiving analytics data.', "In 5GS, the primary role of the Nnef_AnalyticsExposure_Unsubscribe operation is to modify an existing subscription's parameters regarding analytics information.", 'The Nnef_AnalyticsExposure_Unsubscribe service in 5GS is designed to fetch and deliver analytics information to a network function consumer.']</code>

• Loss: <code>MultipleNegativesRankingLoss</code> with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 1,203 evaluation samples
• Columns: <code>query</code>, <code>positive</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	query	positive	negative
  type	string	string	string
  details	<ul><li>min: 14 tokens</li><li>mean: 38.17 tokens</li><li>max: 87 tokens</li></ul>	<ul><li>min: 59 tokens</li><li>mean: 132.72 tokens</li><li>max: 316 tokens</li></ul>	<ul><li>min: 59 tokens</li><li>mean: 115.01 tokens</li><li>max: 241 tokens</li></ul>

• Samples:

  query	positive	negative
  <code>What is the purpose of the Nnsacf_NSAC_NumOfPDUsUpdate service operation in the 5G System (5GS)?</code>	<code>['The Nnsacf_NSAC_NumOfPDUsUpdate service operation in 5GS is designed to modify the count of established PDU Sessions on a specific network slice and verify if the upper limit for PDU Sessions per slice has been met.', 'In 5G System, the purpose of the Nnsacf_NSAC_NumOfPDUsUpdate service is to adjust the number of PDU Sessions on a network slice while ensuring that this count does not surpass the predefined maximum allowed sessions for that slice.', 'The function of the Nnsacf_NSAC_NumOfPDUsUpdate operation in 5GS involves updating the tally of active PDU Sessions within a network slice and checking if it has reached its maximum capacity.']</code>	<code>['The Nnsacf_NSAC_NumOfPDUsUpdate service operation aims to create new PDU Sessions on a network slice without considering any limits on the number of sessions.', 'The role of the Nnsacf_NSAC_NumOfPDUsUpdate is to disconnect existing PDU Sessions from a network slice, ensuring no further sessions are active.', 'In 5GS, the Nnsacf_NSAC_NumOfPDUsUpdate service updates UE ID and PDU Session ID parameters within the NSACF without adjusting session counts.']</code>
  <code>When the Discovery-Type in a Match-Report AVP is set to 'MONITORING_REQUEST_FOR_OPEN_PROSE_DIRECT_DISCOVERY', which of the following elements are MANDATORY to be included in the Match-Report?</code>	<code>["In a Match-Report AVP, when Discovery-Type is 'MONITORING_REQUEST_FOR_OPEN_PROSE_DIRECT_DISCOVERY', it is mandatory to include ProSe-App-Code, ProSe-App-Id, ProSe-Validity-Timer, and ProSe-Match-Refresh-Timer.", "For the discovery type 'MONITORING_REQUEST_FOR_OPEN_PROSE_DIRECT_DISCOVERY' in a Match-Report AVP, required elements are ProSe-App-Code, ProSe-App-Id, ProSe-Validity-Timer, and ProSe-Match-Refresh-Timer.", "When the Discovery-Type is set to 'MONITORING_REQUEST_FOR_OPEN_PROSE_DIRECT_DISCOVERY' in a Match-Report AVP, inclusion of ProSe-App-Code, ProSe-App-Id, ProSe-Validity-Timer, and ProSe-Match-Refresh-Timer is mandatory."]</code>	<code>["In the case where Discovery-Type in a Match-Report AVP is 'MONITORING_REQUEST_FOR_OPEN_PROSE_DIRECT_DISCOVERY', only ProSe-App-Code and PC5-tech are required.", "For a Match-Report with Discovery-Type as 'MONITORING_REQUEST_FOR_OPEN_PROSE_DIRECT_DISCOVERY', it's mandatory to include ProSe-Application-Metadata, ProSe-Validity-Timer, and Discovery-Type itself.", "When the Discovery-Type in a Match-Report AVP is set to 'PROTECTED_SERVICE_ENABLING', elements like ProSe-App-Code, ProSe-App-Id, and ProSe-Match-Refresh-Timer are mandatory."]</code>
  <code>In a 3GPP-compliant FDD UE supporting UL OLTD and HSDPA, when calculating the Nominal CDP ratio for 16QAM modulation, what adjustment is made to the βed value?</code>	<code>['In a 3GPP-compliant FDD user equipment (UE) that supports UL OLTD and HSDPA, what is the adjustment made to the βed value when calculating the Nominal CDP ratio for 16QAM modulation?', 'For 16QAM modulation in an FDD UE with UL OLTD and HSDPA capabilities, how is the βed value adjusted during the calculation of the Nominal CDP ratio?', 'When determining the Nominal CDP ratio for a 3GPP-compliant FDD UE supporting both UL OLTD and HSDPA using 16QAM modulation, what formula adjustment is applied to the βed value?']</code>	<code>['In a 3GPP-compliant FDD UE with UL OLTD and HSDPA, when calculating the Nominal CDP ratio for 16QAM modulation, the βed value is increased by a fixed correction factor of 0.5 dB.', 'For 16QAM modulation in an FDD UE supporting UL OLTD and HSDPA, the adjustment involves multiplying the βed value directly by the sum of all nominal CDPs without any additional operations.', 'When calculating the Nominal CDP ratio for a 3GPP-compliant FDD UE using 16QAM modulation with both UL OLTD and HSDPA, no formula involving symbol fractions is applied to adjust the βed value.']</code>

• Loss: <code>MultipleNegativesRankingLoss</code> with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 5
• warmup_ratio: 0.1
• fp16: True
• batch_sampler: no_duplicates

All Hyperparameters

<details><summary>Click to expand</summary>

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• 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: 1.0
• num_train_epochs: 5
• 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: False
• fp16: True
• 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

</details>

Training Logs

Framework Versions

• Python: 3.10.14
• Sentence Transformers: 3.3.1
• Transformers: 4.49.0
• PyTorch: 2.6.0+cu118
• Accelerate: 1.4.0
• Datasets: 2.21.0
• Tokenizers: 0.21.0

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",
}

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}
}

<!-- ## Glossary *Clearly define terms in order to be accessible across audiences.* --> <!-- ## Model Card Authors *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.* --> <!-- ## Model Card Contact *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.* -->