LucaZilli/arctic-s-phrases-only-v0

SentenceTransformer based on Snowflake/snowflake-arctic-embed-s

This is a sentence-transformers model finetuned from Snowflake/snowflake-arctic-embed-s. It maps sentences & paragraphs to a 384-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: Snowflake/snowflake-arctic-embed-s
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 dimensions
• Similarity Function: Cosine Similarity

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: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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("LucaZilli/arctic-s-phrases-only-v0")
# Run inference
sentences = [
    'CDMO',
    'C.M.L. SNC DI ZANETTI GIOVANNI & C. ||~~|| C.M.L. SNC specializes in precision mechanical machining, offering a range of mechanical processing and various types of machines and systems. mechanical processing',
    'Insight Consulting - Siti web e Digital Marketing ||~~|| Insight Consulting specializes in digital strategy, enhancing customer engagement, brand awareness, and lead acquisition through tailored omni-channel solutions and market analysis. lead acquisition',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

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

Training Details

Training Dataset

Unnamed Dataset

• Size: 48,157 training samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	float
  details	min: 4 tokens mean: 14.6 tokens max: 46 tokens	min: 4 tokens mean: 25.64 tokens max: 133 tokens	min: 0.0 mean: 0.5 max: 1.0

• Samples:

  sentence1	sentence2	score
  pavimentazione industriale antiscivolo certificata	MOLTA SRL
  monitor arm for dual screens	braccio per monitor	0.6
  investigatore privato dipendenti	Investigatore Privato

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 2,352 evaluation samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	float
  details	min: 4 tokens mean: 17.31 tokens max: 48 tokens	min: 13 tokens mean: 71.79 tokens max: 122 tokens	min: 0.0 mean: 0.62 max: 1.0

• Samples:

  sentence1	sentence2	score
  CDMO	C.M.L. SNC DI ZANETTI GIOVANNI & C.
  programmatori salesforce	EFFEGIT SRL
  software con intelligenza artificiale per i contratti	BORRONI VALERIA

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• learning_rate: 2.0000000000000003e-06
• num_train_epochs: 1
• warmup_ratio: 0.1
• fp16: True
• load_best_model_at_end: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• 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: 2.0000000000000003e-06
• 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: True
• 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: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.0066	10	0.1549	-
0.0133	20	0.1566	-
0.0199	30	0.1482	-
0.0266	40	0.135	0.0904
0.0332	50	0.1406	-
0.0399	60	0.1186	-
0.0465	70	0.1077	-
0.0532	80	0.1068	0.0746
0.0598	90	0.0818	-
0.0664	100	0.0841	-
0.0731	110	0.0768	-
0.0797	120	0.0737	0.0721
0.0864	130	0.0717	-
0.0930	140	0.0632	-
0.0997	150	0.0585	-
0.1063	160	0.0633	0.0757
0.1130	170	0.0567	-
0.1196	180	0.0633	-
0.1262	190	0.0584	-
0.1329	200	0.0686	0.0744
0.1395	210	0.0618	-
0.1462	220	0.0585	-
0.1528	230	0.0545	-
0.1595	240	0.0588	0.0721
0.1661	250	0.0532	-
0.1728	260	0.0608	-
0.1794	270	0.054	-
0.1860	280	0.059	0.0697
0.1927	290	0.0513	-
0.1993	300	0.0603	-
0.2060	310	0.0538	-
0.2126	320	0.0565	0.0686
0.2193	330	0.0515	-
0.2259	340	0.0565	-
0.2326	350	0.0579	-
0.2392	360	0.0504	0.0672
0.2458	370	0.0529	-
0.2525	380	0.0541	-
0.2591	390	0.0552	-
0.2658	400	0.0556	0.0669
0.2724	410	0.0561	-
0.2791	420	0.0629	-
0.2857	430	0.05	-
0.2924	440	0.0609	0.0659
0.2990	450	0.0539	-
0.3056	460	0.0556	-
0.3123	470	0.0516	-
0.3189	480	0.0456	0.0651
0.3256	490	0.0485	-
0.3322	500	0.0504	-
0.3389	510	0.0577	-
0.3455	520	0.0538	0.0647
0.3522	530	0.0458	-
0.3588	540	0.0496	-
0.3654	550	0.0486	-
0.3721	560	0.0536	0.0645
0.3787	570	0.0501	-
0.3854	580	0.0519	-
0.3920	590	0.0523	-
0.3987	600	0.0456	0.0639
0.4053	610	0.0561	-
0.4120	620	0.0534	-
0.4186	630	0.0546	-
0.4252	640	0.0531	0.0637
0.4319	650	0.0443	-
0.4385	660	0.0522	-
0.4452	670	0.0456	-
0.4518	680	0.049	0.0635
0.4585	690	0.0488	-
0.4651	700	0.0523	-
0.4718	710	0.0487	-
0.4784	720	0.0515	0.0632
0.4850	730	0.0453	-
0.4917	740	0.0511	-
0.4983	750	0.0429	-
0.5050	760	0.0409	0.0631
0.5116	770	0.0534	-
0.5183	780	0.0485	-
0.5249	790	0.0527	-
0.5316	800	0.0475	0.0630
0.5382	810	0.0512	-
0.5449	820	0.0439	-
0.5515	830	0.042	-
0.5581	840	0.0499	0.0628
0.5648	850	0.0431	-
0.5714	860	0.0541	-
0.5781	870	0.045	-
0.5847	880	0.0495	0.0627
0.5914	890	0.0531	-
0.5980	900	0.0478	-
0.6047	910	0.0547	-
0.6113	920	0.0474	0.0626

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.14
• Sentence Transformers: 3.4.1
• Transformers: 4.49.0
• PyTorch: 2.2.2
• Accelerate: 1.4.0
• Datasets: 3.3.2
• 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",
}