tomaarsen/mpnet-base-nli-matryoshka-reproduced

SentenceTransformer based on microsoft/mpnet-base

This is a sentence-transformers model finetuned from microsoft/mpnet-base on the all-nli 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: microsoft/mpnet-base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • all-nli
• 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: MPNetModel 
  (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/mpnet-base-nli-matryoshka-reproduced")
# Run inference
sentences = [
    'A construction worker peeking out of a manhole while his coworker sits on the sidewalk smiling.',
    'A worker is looking out of a manhole.',
    'The workers are both inside the manhole.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

Evaluation

Metrics

Semantic Similarity

• Datasets: sts-dev-768 and sts-test-768
• Evaluated with EmbeddingSimilarityEvaluator with these parameters:

  {
      "truncate_dim": 768
  }
  

Metric	sts-dev-768	sts-test-768
pearson_cosine	0.8428	0.8189
spearman_cosine	0.8509	0.8359

Semantic Similarity

• Datasets: sts-dev-512 and sts-test-512
• Evaluated with EmbeddingSimilarityEvaluator with these parameters:

  {
      "truncate_dim": 512
  }
  

Metric	sts-dev-512	sts-test-512
pearson_cosine	0.8403	0.8186
spearman_cosine	0.8493	0.8362

Semantic Similarity

• Datasets: sts-dev-256 and sts-test-256
• Evaluated with EmbeddingSimilarityEvaluator with these parameters:

  {
      "truncate_dim": 256
  }
  

Metric	sts-dev-256	sts-test-256
pearson_cosine	0.8347	0.813
spearman_cosine	0.8463	0.8332

Semantic Similarity

• Datasets: sts-dev-128 and sts-test-128
• Evaluated with EmbeddingSimilarityEvaluator with these parameters:

  {
      "truncate_dim": 128
  }
  

Metric	sts-dev-128	sts-test-128
pearson_cosine	0.8258	0.803
spearman_cosine	0.8396	0.8262

Semantic Similarity

• Datasets: sts-dev-64 and sts-test-64
• Evaluated with EmbeddingSimilarityEvaluator with these parameters:

  {
      "truncate_dim": 64
  }
  

Metric	sts-dev-64	sts-test-64
pearson_cosine	0.8134	0.7904
spearman_cosine	0.8314	0.8194

Training Details

Training Dataset

all-nli

• Dataset: all-nli at d482672
• Size: 557,850 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 7 tokens mean: 10.46 tokens max: 46 tokens	min: 6 tokens mean: 12.81 tokens max: 40 tokens	min: 5 tokens mean: 13.4 tokens max: 50 tokens

• Samples:

  anchor	positive	negative
  A person on a horse jumps over a broken down airplane.	A person is outdoors, on a horse.	A person is at a diner, ordering an omelette.
  Children smiling and waving at camera	There are children present	The kids are frowning
  A boy is jumping on skateboard in the middle of a red bridge.	The boy does a skateboarding trick.	The boy skates down the sidewalk.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Evaluation Dataset

all-nli

• Dataset: all-nli at d482672
• Size: 6,584 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 6 tokens mean: 17.95 tokens max: 63 tokens	min: 4 tokens mean: 9.78 tokens max: 29 tokens	min: 5 tokens mean: 10.35 tokens max: 29 tokens

• Samples:

  anchor	positive	negative
  Two women are embracing while holding to go packages.	Two woman are holding packages.	The men are fighting outside a deli.
  Two young children in blue jerseys, one with the number 9 and one with the number 2 are standing on wooden steps in a bathroom and washing their hands in a sink.	Two kids in numbered jerseys wash their hands.	Two kids in jackets walk to school.
  A man selling donuts to a customer during a world exhibition event held in the city of Angeles	A man selling donuts to a customer.	A woman drinks her coffee in a small cafe.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "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: 1
• warmup_ratio: 0.1
• fp16: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• 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: 1
• 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}
• tp_size: 0
• 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
• 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

Training Logs

Epoch	Step	Training Loss	Validation Loss	sts-dev-768_spearman_cosine	sts-dev-512_spearman_cosine	sts-dev-256_spearman_cosine	sts-dev-128_spearman_cosine	sts-dev-64_spearman_cosine	sts-test-768_spearman_cosine	sts-test-512_spearman_cosine	sts-test-256_spearman_cosine	sts-test-128_spearman_cosine	sts-test-64_spearman_cosine
0.0459	1600	4.3243	1.5267	0.8525	0.8475	0.8438	0.8356	0.8155	-	-	-	-	-
0.0918	3200	2.4538	1.4448	0.8479	0.8439	0.8403	0.8346	0.8249	-	-	-	-	-
0.1377	4800	2.2829	1.5117	0.8507	0.8481	0.8429	0.8348	0.8203	-	-	-	-	-
0.1836	6400	2.0446	1.2684	0.8574	0.8541	0.8498	0.8413	0.8302	-	-	-	-	-
0.2294	8000	1.8867	1.3107	0.8452	0.8423	0.8400	0.8352	0.8255	-	-	-	-	-
0.2753	9600	1.747	1.1663	0.8456	0.8420	0.8384	0.8292	0.8229	-	-	-	-	-
0.3212	11200	1.6297	1.0809	0.8420	0.8388	0.8360	0.8294	0.8205	-	-	-	-	-
0.3671	12800	1.5974	1.0853	0.8374	0.8352	0.8310	0.8264	0.8184	-	-	-	-	-
0.4130	14400	1.5227	1.0440	0.8479	0.8457	0.8434	0.8380	0.8266	-	-	-	-	-
0.4589	16000	1.3835	1.0718	0.8365	0.8341	0.8310	0.8258	0.8172	-	-	-	-	-
0.5048	17600	1.3893	1.0140	0.8384	0.8363	0.8339	0.8275	0.8178	-	-	-	-	-
0.5507	19200	1.3203	1.0048	0.8418	0.8400	0.8364	0.8292	0.8204	-	-	-	-	-
0.5966	20800	1.2396	0.9407	0.8458	0.8439	0.8404	0.8353	0.8274	-	-	-	-	-
0.6425	22400	1.1842	0.9541	0.8435	0.8404	0.8384	0.8335	0.8257	-	-	-	-	-
0.6883	24000	1.1217	0.9000	0.8534	0.8512	0.8478	0.8408	0.8297	-	-	-	-	-
0.7342	25600	1.093	0.8731	0.8525	0.8503	0.8467	0.8406	0.8313	-	-	-	-	-
0.7801	27200	1.0609	0.8238	0.8528	0.8510	0.8469	0.8399	0.8312	-	-	-	-	-
0.8260	28800	0.9807	0.8264	0.8497	0.8478	0.8448	0.8384	0.8295	-	-	-	-	-
0.8719	30400	1.0061	0.8135	0.8455	0.8439	0.8405	0.8338	0.8256	-	-	-	-	-
0.9178	32000	0.9724	0.7965	0.8517	0.8499	0.8465	0.8401	0.8319	-	-	-	-	-
0.9637	33600	0.9057	0.7841	0.8509	0.8493	0.8463	0.8396	0.8314	-	-	-	-	-
-1	-1	-	-	-	-	-	-	-	0.8359	0.8362	0.8332	0.8262	0.8194

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Energy Consumed: 0.529 kWh
• Carbon Emitted: 0.206 kg of CO2
• Hours Used: 2.452 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: 4.1.0.dev0
• Transformers: 4.51.1
• PyTorch: 2.6.0+cu124
• Accelerate: 1.5.1
• Datasets: 3.3.2
• Tokenizers: 0.21.1

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}
}