agraharr/telecom-gte-modernbert-matryoshka

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

This is a sentence-transformers model finetuned from BAAI/bge-small-en-v1.5. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: BAAI/bge-small-en-v1.5
• Maximum Sequence Length: 64 tokens
• Output Dimensionality: 384 dimensions
• Similarity Function: Cosine Similarity
• Supported Modality: Text

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'BertModel'})
  (1): Pooling({'embedding_dimension': 384, 'pooling_mode': 'cls', '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("agraharr/telecom-gte-modernbert-matryoshka")
# Run inference
sentences = [
    'What are the key steps involved in the physical-layer processing for the downlink transport channels?',
    'The physical-layer processing for downlink transport channels involves several critical steps. First',
    'The BAP MAPPING CONFIGURATION ACKNOWLEDGE message serves as a response from the gNB-DU to the gNB-CU',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.8686, 0.1211],
#         [0.8686, 1.0000, 0.1872],
#         [0.1211, 0.1872, 1.0000]])

Evaluation

Metrics

Triplet

• Datasets: eval and final
• Evaluated with TripletEvaluator

Metric	eval	final
cosine_accuracy	0.9	0.895

Training Details

Training Dataset

Unnamed Dataset

• Size: 1,800 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 10 tokens mean: 23.42 tokens max: 59 tokens	min: 5 tokens mean: 20.89 tokens max: 64 tokens	min: 2 tokens mean: 21.95 tokens max: 63 tokens

• Samples:

  anchor	positive	negative
  What is the primary purpose of collecting performance measurements in an O-RAN system?	2. To assess the overall performance and capacity of the network.	1. To identify and troubleshoot system failures.
  Which organization is responsible for the development of the O-RAN specifications?	3. O-RAN ALLIANCE	1. 3GPP
  What are the capabilities of a UE in Carrier Aggregation when it comes to timing advance?	In Carrier Aggregation (CA), a User Equipment (UE) can operate under different timing advance capabi	The PUSCH Pathloss Reference RS Update MAC CE includes several key fields: The Serving Cell ID, whic

• Loss: MatryoshkaLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 200 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 200 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 14 tokens mean: 25.21 tokens max: 47 tokens	min: 5 tokens mean: 21.38 tokens max: 50 tokens	min: 4 tokens mean: 23.97 tokens max: 62 tokens

• Samples:

  anchor	positive	negative
  What is the maximum acceptable relative time error between the O-DU and O-RU for S-plane measurement signals in O-RAN, according to the provided context?	2. 3 µs	1. 1.5 µs
  What are the calibration metrics calibrated for BS antenna configuration Config 2 in Table 7.8-2?	Table 7.8-2: Simulation assumptions for full calibration	The UE variable VarRA-Report includes the random-access related information. VarRA-Report UE variable -- ASN1START -- TAG-VARRA-REPORT-START VarRA-Rep
  How does the scheduled modulation order affect the PT-RS scaling factor when transform precoding is enabled?	When transform precoding is enabled, the PT-RS scaling factor (β') is determined based on the schedu	The spurious response test for 2DL carrier aggregation (CA) is designed to assess the receiver's cap

• Loss: MatryoshkaLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 4
• num_train_epochs: 2
• learning_rate: 3e-05
• warmup_steps: 0.1
• disable_tqdm: True
• per_device_eval_batch_size: 4
• push_to_hub: True
• hub_model_id: agraharr/telecom-gte-modernbert-matryoshka
• hub_strategy: end
• load_best_model_at_end: True
• dataloader_pin_memory: False

All Hyperparameters

Click to expand

• per_device_train_batch_size: 4
• num_train_epochs: 2
• max_steps: -1
• learning_rate: 3e-05
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_steps: 0.1
• optim: adamw_torch_fused
• optim_args: None
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• optim_target_modules: None
• gradient_accumulation_steps: 1
• average_tokens_across_devices: True
• max_grad_norm: 1.0
• label_smoothing_factor: 0.0
• bf16: False
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• use_liger_kernel: False
• liger_kernel_config: None
• use_cache: False
• neftune_noise_alpha: None
• torch_empty_cache_steps: None
• auto_find_batch_size: False
• log_on_each_node: True
• logging_nan_inf_filter: True
• include_num_input_tokens_seen: no
• log_level: passive
• log_level_replica: warning
• disable_tqdm: True
• project: huggingface
• trackio_space_id: None
• trackio_bucket_id: None
• trackio_static_space_id: None
• per_device_eval_batch_size: 4
• prediction_loss_only: True
• eval_on_start: False
• eval_do_concat_batches: True
• eval_use_gather_object: False
• eval_accumulation_steps: None
• include_for_metrics: []
• batch_eval_metrics: False
• save_only_model: False
• save_on_each_node: False
• enable_jit_checkpoint: False
• push_to_hub: True
• hub_private_repo: None
• hub_model_id: agraharr/telecom-gte-modernbert-matryoshka
• hub_strategy: end
• hub_always_push: False
• hub_revision: None
• load_best_model_at_end: True
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• data_seed: None
• use_cpu: False
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• parallelism_config: None
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_pin_memory: False
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• remove_unused_columns: True
• label_names: None
• train_sampling_strategy: random
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• ddp_static_graph: None
• ddp_backend: None
• ddp_timeout: 1800
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• deepspeed: None
• debug: []
• skip_memory_metrics: True
• do_predict: False
• resume_from_checkpoint: None
• warmup_ratio: None
• local_rank: -1
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	eval_cosine_accuracy	final_cosine_accuracy
-1	-1	-	-	0.8000	-
0.0022	1	1.2819	-	-	-
0.1111	50	2.1454	-	-	-
0.2222	100	1.2912	-	-	-
0.3333	150	1.3769	-	-	-
0.4444	200	1.5782	-	-	-
0.5556	250	1.0937	-	-	-
0.6667	300	1.0673	-	-	-
0.7778	350	1.2251	-	-	-
0.8889	400	1.0413	-	-	-
1.0	450	0.8361	0.8982	0.8800	-
1.1111	500	0.6237	-	-	-
1.2222	550	0.7264	-	-	-
1.3333	600	0.5985	-	-	-
1.4444	650	0.7544	-	-	-
1.5556	700	0.7694	-	-	-
1.6667	750	0.6571	-	-	-
1.7778	800	0.4875	-	-	-
1.8889	850	0.5598	-	-	-
2.0	900	0.5807	0.8917	0.9	-
-1	-1	-	-	-	0.8950

• The bold row denotes the saved checkpoint.

Training Time

• Training: 13.8 minutes
• Evaluation: 28.7 seconds
• Total: 14.3 minutes

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.4.1
• Transformers: 5.6.1
• PyTorch: 2.11.0+cu130
• Accelerate: 1.13.0
• Datasets: 4.8.4
• Tokenizers: 0.22.2

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{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}

Generated by ML Intern

This model repository was generated by ML Intern, an agent for machine learning research and development on the Hugging Face Hub.

• Try ML Intern: https://smolagents-ml-intern.hf.space
• Source code: https://github.com/huggingface/ml-intern