tomaarsen/distilbert-base-uncased-sts-2d-matryoshka

SentenceTransformer based on distilbert/distilbert-base-uncased

This is a sentence-transformers model finetuned from distilbert/distilbert-base-uncased on the sentence-transformers/stsb dataset. 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
• Base model: distilbert/distilbert-base-uncased
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity
• Training Dataset:
  • sentence-transformers/stsb
• Language: en

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: DistilBertModel 
  (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("tomaarsen/distilbert-base-uncased-sts-2d-matryoshka")
# Run inference
sentences = [
    'A parrot is talking.',
    'A parrot is talking into a microphone.',
    'A monkey pratices martial arts.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Semantic Similarity

• Dataset: sts-dev
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8619
spearman_cosine	0.8713
pearson_manhattan	0.8611
spearman_manhattan	0.8619
pearson_euclidean	0.8615
spearman_euclidean	0.8624
pearson_dot	0.7716
spearman_dot	0.7816
pearson_max	0.8619
spearman_max	0.8713

Semantic Similarity

• Dataset: sts-test
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8282
spearman_cosine	0.8373
pearson_manhattan	0.8425
spearman_manhattan	0.8374
pearson_euclidean	0.8425
spearman_euclidean	0.8373
pearson_dot	0.675
spearman_dot	0.6568
pearson_max	0.8425
spearman_max	0.8374

Training Details

Training Dataset

sentence-transformers/stsb

• Dataset: sentence-transformers/stsb at ab7a5ac
• Size: 5,749 training samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	float
  details	min: 6 tokens mean: 10.0 tokens max: 28 tokens	min: 5 tokens mean: 9.95 tokens max: 25 tokens	min: 0.0 mean: 0.54 max: 1.0

• Samples:

  sentence1	sentence2	score
  A plane is taking off.	An air plane is taking off.	1.0
  A man is playing a large flute.	A man is playing a flute.	0.76
  A man is spreading shreded cheese on a pizza.	A man is spreading shredded cheese on an uncooked pizza.	0.76

• Loss: Matryoshka2dLoss with these parameters:

  {
      "loss": "CoSENTLoss",
      "n_layers_per_step": 1,
      "last_layer_weight": 1.0,
      "prior_layers_weight": 1.0,
      "kl_div_weight": 1.0,
      "kl_temperature": 0.3,
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": 1
  }
  

Evaluation Dataset

sentence-transformers/stsb

• Dataset: sentence-transformers/stsb at ab7a5ac
• Size: 1,500 evaluation samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	float
  details	min: 5 tokens mean: 15.1 tokens max: 45 tokens	min: 6 tokens mean: 15.11 tokens max: 53 tokens	min: 0.0 mean: 0.47 max: 1.0

• Samples:

  sentence1	sentence2	score
  A man with a hard hat is dancing.	A man wearing a hard hat is dancing.	1.0
  A young child is riding a horse.	A child is riding a horse.	0.95
  A man is feeding a mouse to a snake.	The man is feeding a mouse to the snake.	1.0

• Loss: Matryoshka2dLoss with these parameters:

  {
      "loss": "CoSENTLoss",
      "n_layers_per_step": 1,
      "last_layer_weight": 1.0,
      "prior_layers_weight": 1.0,
      "kl_div_weight": 1.0,
      "kl_temperature": 0.3,
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": 1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 4
• warmup_ratio: 0.1
• fp16: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: False
• 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
• 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: 4
• 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
• 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: None
• 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_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	sts-dev_spearman_cosine	sts-test_spearman_cosine
0.2778	100	7.1781	6.6704	0.8345	-
0.5556	200	6.5316	6.7135	0.8439	-
0.8333	300	6.6267	6.8697	0.8551	-
1.1111	400	6.5709	6.7623	0.8568	-
1.3889	500	6.2898	6.4412	0.8644	-
1.6667	600	6.2021	6.7711	0.8595	-
1.9444	700	6.201	6.5252	0.8628	-
2.2222	800	6.0862	6.9795	0.8652	-
2.5	900	6.303	6.7339	0.8685	-
2.7778	1000	5.9031	6.7249	0.8694	-
3.0556	1100	6.0803	6.8350	0.8684	-
3.3333	1200	6.0564	6.9703	0.8695	-
3.6111	1300	5.8407	7.3822	0.8707	-
3.8889	1400	5.8229	7.0442	0.8713	-
4.0	1440	-	-	-	0.8373

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Energy Consumed: 0.014 kWh
• Carbon Emitted: 0.005 kg of CO2
• Hours Used: 0.072 hours

Training Hardware

• On Cloud: No
• GPU Model: 1 x NVIDIA GeForce RTX 3090
• CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
• RAM Size: 31.78 GB

Framework Versions

• Python: 3.11.6
• Sentence Transformers: 3.0.0.dev0
• Transformers: 4.41.0.dev0
• PyTorch: 2.3.0+cu121
• Accelerate: 0.26.1
• Datasets: 2.18.0
• 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",
}

Matryoshka2dLoss

@misc{li20242d,
    title={2D Matryoshka Sentence Embeddings}, 
    author={Xianming Li and Zongxi Li and Jing Li and Haoran Xie and Qing Li},
    year={2024},
    eprint={2402.14776},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

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}
}

CoSENTLoss

@online{kexuefm-8847,
    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
    author={Su Jianlin},
    year={2022},
    month={Jan},
    url={https://kexue.fm/archives/8847},
}