SentenceTransformer based on distilbert/distilroberta-base

This is a sentence-transformers model finetuned from distilbert/distilroberta-base on the sentence-transformers/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: distilbert/distilroberta-base
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 tokens
Similarity Function: Cosine Similarity
Training Dataset:
- sentence-transformers/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: RobertaModel 
  (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("mrm8488/distilroberta-base-ft-allnli-matryoshka-768-16-1e-128bs")
# Run inference
sentences = [
    'It is well.',
    "That's convenient.",
    'away from the children',
]
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

Dataset: sts-dev-768
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8413
spearman_cosine	0.8478
pearson_manhattan	0.8414
spearman_manhattan	0.8395
pearson_euclidean	0.8423
spearman_euclidean	0.8401
pearson_dot	0.7855
spearman_dot	0.7814
pearson_max	0.8423
spearman_max	0.8478

Semantic Similarity

Dataset: sts-dev-512
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8395
spearman_cosine	0.847
pearson_manhattan	0.8399
spearman_manhattan	0.8377
pearson_euclidean	0.8407
spearman_euclidean	0.838
pearson_dot	0.7811
spearman_dot	0.7777
pearson_max	0.8407
spearman_max	0.847

Semantic Similarity

Dataset: sts-dev-256
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8327
spearman_cosine	0.8436
pearson_manhattan	0.8351
spearman_manhattan	0.8332
pearson_euclidean	0.836
spearman_euclidean	0.8338
pearson_dot	0.75
spearman_dot	0.7453
pearson_max	0.836
spearman_max	0.8436

Semantic Similarity

Dataset: sts-dev-128
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8243
spearman_cosine	0.84
pearson_manhattan	0.8282
spearman_manhattan	0.827
pearson_euclidean	0.8282
spearman_euclidean	0.8267
pearson_dot	0.711
spearman_dot	0.705
pearson_max	0.8282
spearman_max	0.84

Semantic Similarity

Dataset: sts-dev-64
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8116
spearman_cosine	0.8317
pearson_manhattan	0.8113
spearman_manhattan	0.8105
pearson_euclidean	0.8114
spearman_euclidean	0.8111
pearson_dot	0.6412
spearman_dot	0.6347
pearson_max	0.8116
spearman_max	0.8317

Semantic Similarity

Dataset: sts-dev-32
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.7834
spearman_cosine	0.8141
pearson_manhattan	0.7832
spearman_manhattan	0.786
pearson_euclidean	0.7869
spearman_euclidean	0.7894
pearson_dot	0.5534
spearman_dot	0.5449
pearson_max	0.7869
spearman_max	0.8141

Semantic Similarity

Dataset: sts-dev-16
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.7259
spearman_cosine	0.7751
pearson_manhattan	0.7421
spearman_manhattan	0.7553
pearson_euclidean	0.7483
spearman_euclidean	0.7599
pearson_dot	0.4387
spearman_dot	0.4259
pearson_max	0.7483
spearman_max	0.7751

Training Details

Training Dataset

sentence-transformers/all-nli

Dataset: sentence-transformers/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: 10.38 tokens max: 45 tokens	min: 6 tokens mean: 12.8 tokens max: 39 tokens	min: 6 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,
        32,
        16
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Evaluation Dataset

sentence-transformers/all-nli

Dataset: sentence-transformers/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: 18.02 tokens max: 66 tokens	min: 5 tokens mean: 9.81 tokens max: 29 tokens	min: 5 tokens mean: 10.37 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,
        32,
        16
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
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: 128
per_device_eval_batch_size: 128
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: 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}
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: 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_eval_metrics: False
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	sts-dev-128_spearman_cosine	sts-dev-16_spearman_cosine	sts-dev-256_spearman_cosine	sts-dev-32_spearman_cosine	sts-dev-512_spearman_cosine	sts-dev-64_spearman_cosine	sts-dev-768_spearman_cosine
0.0229	100	29.0917	14.1514	0.7659	0.7440	0.7915	0.7749	0.7999	0.7909	0.7918
0.0459	200	15.6915	11.7031	0.7718	0.7487	0.7940	0.7776	0.8005	0.7931	0.7871
0.0688	300	14.3136	11.1970	0.7744	0.7389	0.7952	0.7728	0.8036	0.7925	0.7938
0.0918	400	12.8122	10.4416	0.7899	0.7536	0.8040	0.7764	0.8065	0.7953	0.8018
0.1147	500	12.1747	10.5491	0.7871	0.7513	0.8035	0.7785	0.8094	0.7978	0.8008
0.1376	600	11.6784	9.6618	0.7785	0.7465	0.7956	0.7762	0.8027	0.7953	0.7935
0.1606	700	11.9351	9.3279	0.7907	0.7403	0.7995	0.7706	0.8036	0.7894	0.7982
0.1835	800	10.4998	9.1538	0.7911	0.7516	0.8043	0.7820	0.8078	0.8025	0.8010
0.2065	900	10.6069	9.0531	0.7874	0.7371	0.7974	0.7704	0.8042	0.7910	0.8010
0.2294	1000	10.0316	8.9759	0.7842	0.7356	0.7981	0.7721	0.8024	0.7905	0.7955
0.2524	1100	10.199	8.5398	0.7863	0.7322	0.7961	0.7691	0.8002	0.7910	0.7936
0.2753	1200	9.9393	8.1356	0.7860	0.7304	0.7990	0.7682	0.8025	0.7908	0.7954
0.2982	1300	9.8711	7.9177	0.7932	0.7319	0.8028	0.7708	0.8067	0.7924	0.8013
0.3212	1400	9.3594	7.8870	0.7892	0.7296	0.8032	0.7710	0.8070	0.7961	0.8030
0.3441	1500	9.4534	7.5756	0.8003	0.7518	0.8078	0.7857	0.8112	0.8063	0.8068
0.3671	1600	8.9061	7.8164	0.7781	0.7390	0.7942	0.7761	0.8002	0.7968	0.7941
0.3900	1700	8.5164	7.4869	0.7934	0.7530	0.8063	0.7864	0.8120	0.8055	0.8080
0.4129	1800	8.9262	7.7155	0.7846	0.7301	0.7991	0.7728	0.8065	0.7945	0.8003
0.4359	1900	8.3242	7.3068	0.7850	0.7273	0.7976	0.7710	0.8020	0.7904	0.7976
0.4588	2000	8.5374	7.1026	0.7845	0.7272	0.7993	0.7717	0.8042	0.7925	0.7963
0.4818	2100	8.2304	7.1601	0.7879	0.7354	0.8015	0.7719	0.8059	0.7944	0.8029
0.5047	2200	8.1347	7.8267	0.7715	0.7230	0.7889	0.7626	0.7956	0.7849	0.7930
0.5276	2300	8.3057	8.0057	0.7622	0.7148	0.7814	0.7572	0.7881	0.7769	0.7836
0.5506	2400	8.215	7.6922	0.7772	0.7210	0.7929	0.7637	0.7995	0.7858	0.7956
0.5735	2500	8.4343	7.2104	0.7869	0.7307	0.8017	0.7707	0.8071	0.7929	0.8048
0.5965	2600	8.159	6.9977	0.7893	0.7297	0.8031	0.7733	0.8071	0.7928	0.8045
0.6194	2700	8.2048	6.9465	0.7859	0.7280	0.8006	0.7725	0.8052	0.7926	0.8004
0.6423	2800	8.187	7.3185	0.7790	0.7266	0.7960	0.7690	0.8018	0.7911	0.7964
0.6653	2900	8.4768	7.5535	0.7756	0.7192	0.7913	0.7618	0.7958	0.7827	0.7907
0.6882	3000	8.4153	7.3732	0.7825	0.7276	0.7988	0.7692	0.8029	0.7899	0.7988
0.7112	3100	7.9226	6.8469	0.7912	0.7311	0.8055	0.7765	0.8101	0.7977	0.8058
0.7341	3200	8.1155	6.7604	0.7880	0.7298	0.8024	0.7747	0.8071	0.7959	0.8025
0.7571	3300	6.8463	5.4863	0.8357	0.7638	0.8407	0.8085	0.8431	0.8283	0.8440
0.7800	3400	5.2008	5.2472	0.8362	0.7655	0.8401	0.8105	0.8429	0.8279	0.8445
0.8029	3500	4.5415	5.1649	0.8385	0.7700	0.8421	0.8138	0.8454	0.8304	0.8465
0.8259	3600	4.4474	5.0933	0.8371	0.7693	0.8410	0.8112	0.8443	0.8288	0.8451
0.8488	3700	4.12	5.0555	0.8396	0.7718	0.8439	0.8140	0.8463	0.8311	0.8471
0.8718	3800	3.9104	5.0147	0.8386	0.7749	0.8432	0.8129	0.8459	0.8304	0.8471
0.8947	3900	3.9054	4.9966	0.8379	0.7733	0.8424	0.8125	0.8456	0.8296	0.8464
0.9176	4000	3.757	4.9892	0.8407	0.7763	0.8447	0.8156	0.8478	0.8326	0.8488
0.9406	4100	3.7729	4.9859	0.8400	0.7751	0.8436	0.8141	0.8470	0.8317	0.8478

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.0
Transformers: 4.41.1
PyTorch: 2.3.0+cu121
Accelerate: 0.30.1
Datasets: 2.19.2
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",
}

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}
}

Downloads last month: -

Safetensors

Model size

82.1M params

Tensor type

F32

Model tree for mrm8488/distilroberta-base-ft-allnli-matryoshka-768-16-1e-128bs

Base model

distilbert/distilroberta-base

Finetuned

(708)

this model

Evaluation results

Pearson Cosine on sts dev 768
self-reported

0.841
Spearman Cosine on sts dev 768
self-reported

0.848
Pearson Manhattan on sts dev 768
self-reported

0.841
Spearman Manhattan on sts dev 768
self-reported

0.839
Pearson Euclidean on sts dev 768
self-reported

0.842
Spearman Euclidean on sts dev 768
self-reported

0.840
Pearson Dot on sts dev 768
self-reported

0.785
Spearman Dot on sts dev 768
self-reported

0.781
Pearson Max on sts dev 768
self-reported

0.842
Spearman Max on sts dev 768
self-reported

0.848

View on Papers With Code