SenhorDasMoscas/bert-ptbr-e3-lr0.0001-04-01-2025

SentenceTransformer based on neuralmind/bert-large-portuguese-cased

This is a sentence-transformers model finetuned from neuralmind/bert-large-portuguese-cased. It maps sentences & paragraphs to a 1024-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: neuralmind/bert-large-portuguese-cased
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 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': 1024, '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("SenhorDasMoscas/bert-ptbr-e3-lr0.0001-04-01-2025")
# Run inference
sentences = [
    'cobertor pelucia',
    'moda acessorio',
    'servico reparo eletronico',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Evaluation

Metrics

Semantic Similarity

• Dataset: eval-similarity
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9058
spearman_cosine	0.8399

Training Details

Training Dataset

Unnamed Dataset

• Size: 18,623 training samples
• Columns: text1, text2, and label
• Approximate statistics based on the first 1000 samples:

  	text1	text2	label
  type	string	string	float
  details	min: 3 tokens mean: 7.67 tokens max: 17 tokens	min: 3 tokens mean: 6.58 tokens max: 11 tokens	min: 0.1 mean: 0.54 max: 1.0

• Samples:

  text1	text2	label
  tabua carne	casa decoracao	1.0
  caminhaor basculante brinquedo	brinquedo jogo educativo	1.0
  buscar mochila escolar crianca	comida rapido fastfood	0.1

• Loss: CosineSimilarityLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 2,070 evaluation samples
• Columns: text1, text2, and label
• Approximate statistics based on the first 1000 samples:

  	text1	text2	label
  type	string	string	float
  details	min: 3 tokens mean: 7.69 tokens max: 17 tokens	min: 3 tokens mean: 6.54 tokens max: 11 tokens	min: 0.1 mean: 0.59 max: 1.0

• Samples:

  text1	text2	label
  preciso pao frances integral	padaria confeitaria	1.0
  onde poder comprar microfone	joia bijuterio	0.1
  chuveiro eletrico lorenzetti	livro material literario	0.1

• 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: 0.0001
• weight_decay: 0.1
• warmup_ratio: 0.1
• warmup_steps: 232
• 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: 0.0001
• weight_decay: 0.1
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 3
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 232
• 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

Click to expand

Epoch	Step	Training Loss	Validation Loss	eval-similarity_spearman_cosine
0.0086	5	0.2031	-	-
0.0172	10	0.2078	-	-
0.0258	15	0.2062	-	-
0.0344	20	0.1693	-	-
0.0430	25	0.1681	-	-
0.0515	30	0.1639	-	-
0.0601	35	0.1393	-	-
0.0687	40	0.1675	-	-
0.0773	45	0.1297	-	-
0.0859	50	0.1223	-	-
0.0945	55	0.1203	-	-
0.1031	60	0.0942	-	-
0.1117	65	0.0922	-	-
0.1203	70	0.097	-	-
0.1289	75	0.0927	-	-
0.1375	80	0.0961	-	-
0.1460	85	0.0821	-	-
0.1546	90	0.0621	-	-
0.1632	95	0.084	-	-
0.1718	100	0.0706	-	-
0.1804	105	0.0701	-	-
0.1890	110	0.0828	-	-
0.1976	115	0.078	-	-
0.2062	120	0.0745	-	-
0.2148	125	0.0744	-	-
0.2234	130	0.0785	-	-
0.2320	135	0.0745	-	-
0.2405	140	0.0615	-	-
0.2491	145	0.0665	-	-
0.2577	150	0.0873	-	-
0.2663	155	0.0916	-	-
0.2749	160	0.0659	-	-
0.2835	165	0.0896	-	-
0.2921	170	0.0807	-	-
0.3007	175	0.0745	-	-
0.3093	180	0.0794	-	-
0.3179	185	0.0703	-	-
0.3265	190	0.0705	-	-
0.3351	195	0.084	-	-
0.3436	200	0.0671	-	-
0.3522	205	0.076	-	-
0.3608	210	0.0821	-	-
0.3694	215	0.0499	-	-
0.3780	220	0.0729	-	-
0.3866	225	0.0697	-	-
0.3952	230	0.085	-	-
0.4038	235	0.0835	-	-
0.4124	240	0.0743	-	-
0.4210	245	0.0714	-	-
0.4296	250	0.0597	-	-
0.4381	255	0.0626	-	-
0.4467	260	0.0522	-	-
0.4553	265	0.0734	-	-
0.4639	270	0.0616	-	-
0.4725	275	0.0463	-	-
0.4811	280	0.0631	-	-
0.4897	285	0.0672	-	-
0.4983	290	0.0725	-	-
0.5069	295	0.043	-	-
0.5155	300	0.0675	0.0698	0.7861
0.5241	305	0.0837	-	-
0.5326	310	0.0785	-	-
0.5412	315	0.0761	-	-
0.5498	320	0.0523	-	-
0.5584	325	0.0514	-	-
0.5670	330	0.0726	-	-
0.5756	335	0.0584	-	-
0.5842	340	0.0736	-	-
0.5928	345	0.0705	-	-
0.6014	350	0.0682	-	-
0.6100	355	0.0636	-	-
0.6186	360	0.0484	-	-
0.6271	365	0.0524	-	-
0.6357	370	0.0657	-	-
0.6443	375	0.0766	-	-
0.6529	380	0.0759	-	-
0.6615	385	0.071	-	-
0.6701	390	0.055	-	-
0.6787	395	0.0466	-	-
0.6873	400	0.0697	-	-
0.6959	405	0.0546	-	-
0.7045	410	0.0692	-	-
0.7131	415	0.0519	-	-
0.7216	420	0.0521	-	-
0.7302	425	0.0449	-	-
0.7388	430	0.0646	-	-
0.7474	435	0.0585	-	-
0.7560	440	0.0536	-	-
0.7646	445	0.0592	-	-
0.7732	450	0.0515	-	-
0.7818	455	0.0676	-	-
0.7904	460	0.0732	-	-
0.7990	465	0.0618	-	-
0.8076	470	0.0579	-	-
0.8162	475	0.0516	-	-
0.8247	480	0.0659	-	-
0.8333	485	0.0583	-	-
0.8419	490	0.0624	-	-
0.8505	495	0.0667	-	-
0.8591	500	0.052	-	-
0.8677	505	0.0858	-	-
0.8763	510	0.0441	-	-
0.8849	515	0.0592	-	-
0.8935	520	0.0532	-	-
0.9021	525	0.0478	-	-
0.9107	530	0.062	-	-
0.9192	535	0.0487	-	-
0.9278	540	0.0704	-	-
0.9364	545	0.0467	-	-
0.9450	550	0.0482	-	-
0.9536	555	0.0796	-	-
0.9622	560	0.0568	-	-
0.9708	565	0.0588	-	-
0.9794	570	0.0514	-	-
0.9880	575	0.0543	-	-
0.9966	580	0.0568	-	-
1.0052	585	0.0513	-	-
1.0137	590	0.0361	-	-
1.0223	595	0.0405	-	-
1.0309	600	0.0347	0.0491	0.8180
1.0395	605	0.0459	-	-
1.0481	610	0.0557	-	-
1.0567	615	0.0447	-	-
1.0653	620	0.0279	-	-
1.0739	625	0.0417	-	-
1.0825	630	0.025	-	-
1.0911	635	0.0399	-	-
1.0997	640	0.0466	-	-
1.1082	645	0.0294	-	-
1.1168	650	0.035	-	-
1.1254	655	0.0376	-	-
1.1340	660	0.0414	-	-
1.1426	665	0.0502	-	-
1.1512	670	0.04	-	-
1.1598	675	0.0385	-	-
1.1684	680	0.0286	-	-
1.1770	685	0.0361	-	-
1.1856	690	0.0282	-	-
1.1942	695	0.0473	-	-
1.2027	700	0.0346	-	-
1.2113	705	0.0295	-	-
1.2199	710	0.0283	-	-
1.2285	715	0.0301	-	-
1.2371	720	0.0565	-	-
1.2457	725	0.0325	-	-
1.2543	730	0.0299	-	-
1.2629	735	0.0417	-	-
1.2715	740	0.0398	-	-
1.2801	745	0.0477	-	-
1.2887	750	0.0418	-	-
1.2973	755	0.034	-	-
1.3058	760	0.0397	-	-
1.3144	765	0.0308	-	-
1.3230	770	0.0457	-	-
1.3316	775	0.0328	-	-
1.3402	780	0.0222	-	-
1.3488	785	0.0246	-	-
1.3574	790	0.0229	-	-
1.3660	795	0.0351	-	-
1.3746	800	0.0415	-	-
1.3832	805	0.0351	-	-
1.3918	810	0.0269	-	-
1.4003	815	0.0307	-	-
1.4089	820	0.0381	-	-
1.4175	825	0.0425	-	-
1.4261	830	0.0557	-	-
1.4347	835	0.0523	-	-
1.4433	840	0.0488	-	-
1.4519	845	0.0355	-	-
1.4605	850	0.0403	-	-
1.4691	855	0.0332	-	-
1.4777	860	0.0427	-	-
1.4863	865	0.0348	-	-
1.4948	870	0.0375	-	-
1.5034	875	0.0271	-	-
1.5120	880	0.0428	-	-
1.5206	885	0.0666	-	-
1.5292	890	0.0491	-	-
1.5378	895	0.0424	-	-
1.5464	900	0.0413	0.0418	0.8326
1.5550	905	0.0469	-	-
1.5636	910	0.0288	-	-
1.5722	915	0.0541	-	-
1.5808	920	0.017	-	-
1.5893	925	0.0505	-	-
1.5979	930	0.0341	-	-
1.6065	935	0.0223	-	-
1.6151	940	0.0469	-	-
1.6237	945	0.0386	-	-
1.6323	950	0.0214	-	-
1.6409	955	0.0329	-	-
1.6495	960	0.0398	-	-
1.6581	965	0.0355	-	-
1.6667	970	0.0373	-	-
1.6753	975	0.0339	-	-
1.6838	980	0.0349	-	-
1.6924	985	0.0439	-	-
1.7010	990	0.0425	-	-
1.7096	995	0.0318	-	-
1.7182	1000	0.025	-	-
1.7268	1005	0.0334	-	-
1.7354	1010	0.0327	-	-
1.7440	1015	0.0356	-	-
1.7526	1020	0.0428	-	-
1.7612	1025	0.0432	-	-
1.7698	1030	0.0334	-	-
1.7784	1035	0.032	-	-
1.7869	1040	0.0318	-	-
1.7955	1045	0.0281	-	-
1.8041	1050	0.0231	-	-
1.8127	1055	0.0436	-	-
1.8213	1060	0.0303	-	-
1.8299	1065	0.0489	-	-
1.8385	1070	0.0292	-	-
1.8471	1075	0.06	-	-
1.8557	1080	0.0329	-	-
1.8643	1085	0.0322	-	-
1.8729	1090	0.0426	-	-
1.8814	1095	0.0263	-	-
1.8900	1100	0.024	-	-
1.8986	1105	0.0228	-	-
1.9072	1110	0.0313	-	-
1.9158	1115	0.044	-	-
1.9244	1120	0.036	-	-
1.9330	1125	0.0252	-	-
1.9416	1130	0.0311	-	-
1.9502	1135	0.0452	-	-
1.9588	1140	0.0338	-	-
1.9674	1145	0.0447	-	-
1.9759	1150	0.0318	-	-
1.9845	1155	0.0428	-	-
1.9931	1160	0.03	-	-
2.0017	1165	0.0314	-	-
2.0103	1170	0.0181	-	-
2.0189	1175	0.0137	-	-
2.0275	1180	0.0242	-	-
2.0361	1185	0.03	-	-
2.0447	1190	0.0267	-	-
2.0533	1195	0.0263	-	-
2.0619	1200	0.0219	0.0392	0.8360
2.0704	1205	0.0189	-	-
2.0790	1210	0.0193	-	-
2.0876	1215	0.0345	-	-
2.0962	1220	0.0136	-	-
2.1048	1225	0.0346	-	-
2.1134	1230	0.0163	-	-
2.1220	1235	0.0264	-	-
2.1306	1240	0.0172	-	-
2.1392	1245	0.0163	-	-
2.1478	1250	0.0226	-	-
2.1564	1255	0.0229	-	-
2.1649	1260	0.0185	-	-
2.1735	1265	0.0134	-	-
2.1821	1270	0.0144	-	-
2.1907	1275	0.0215	-	-
2.1993	1280	0.0291	-	-
2.2079	1285	0.0305	-	-
2.2165	1290	0.0192	-	-
2.2251	1295	0.0272	-	-
2.2337	1300	0.0267	-	-
2.2423	1305	0.0265	-	-
2.2509	1310	0.0207	-	-
2.2595	1315	0.0305	-	-
2.2680	1320	0.0292	-	-
2.2766	1325	0.017	-	-
2.2852	1330	0.0242	-	-
2.2938	1335	0.016	-	-
2.3024	1340	0.0241	-	-
2.3110	1345	0.0193	-	-
2.3196	1350	0.0134	-	-
2.3282	1355	0.0206	-	-
2.3368	1360	0.0218	-	-
2.3454	1365	0.0239	-	-
2.3540	1370	0.0314	-	-
2.3625	1375	0.028	-	-
2.3711	1380	0.021	-	-
2.3797	1385	0.0179	-	-
2.3883	1390	0.0173	-	-
2.3969	1395	0.0228	-	-
2.4055	1400	0.0217	-	-
2.4141	1405	0.0243	-	-
2.4227	1410	0.018	-	-
2.4313	1415	0.0233	-	-
2.4399	1420	0.016	-	-
2.4485	1425	0.0308	-	-
2.4570	1430	0.0239	-	-
2.4656	1435	0.018	-	-
2.4742	1440	0.016	-	-
2.4828	1445	0.0189	-	-
2.4914	1450	0.0215	-	-
2.5	1455	0.027	-	-
2.5086	1460	0.0177	-	-
2.5172	1465	0.0325	-	-
2.5258	1470	0.0136	-	-
2.5344	1475	0.0235	-	-
2.5430	1480	0.0362	-	-
2.5515	1485	0.0302	-	-
2.5601	1490	0.0137	-	-
2.5687	1495	0.0162	-	-
2.5773	1500	0.0174	0.0376	0.8399
2.5859	1505	0.0248	-	-
2.5945	1510	0.0131	-	-
2.6031	1515	0.0188	-	-
2.6117	1520	0.011	-	-
2.6203	1525	0.0174	-	-
2.6289	1530	0.0192	-	-
2.6375	1535	0.0113	-	-
2.6460	1540	0.0304	-	-
2.6546	1545	0.0217	-	-
2.6632	1550	0.0102	-	-
2.6718	1555	0.0164	-	-
2.6804	1560	0.017	-	-
2.6890	1565	0.0146	-	-
2.6976	1570	0.0139	-	-
2.7062	1575	0.0171	-	-
2.7148	1580	0.0137	-	-
2.7234	1585	0.008	-	-
2.7320	1590	0.0222	-	-
2.7405	1595	0.0295	-	-
2.7491	1600	0.0178	-	-
2.7577	1605	0.0144	-	-
2.7663	1610	0.023	-	-
2.7749	1615	0.0135	-	-
2.7835	1620	0.0213	-	-
2.7921	1625	0.0213	-	-
2.8007	1630	0.0212	-	-
2.8093	1635	0.0164	-	-
2.8179	1640	0.0212	-	-
2.8265	1645	0.0157	-	-
2.8351	1650	0.0251	-	-
2.8436	1655	0.0276	-	-
2.8522	1660	0.0104	-	-
2.8608	1665	0.0123	-	-
2.8694	1670	0.0339	-	-
2.8780	1675	0.0203	-	-
2.8866	1680	0.0171	-	-
2.8952	1685	0.0304	-	-
2.9038	1690	0.015	-	-
2.9124	1695	0.0177	-	-
2.9210	1700	0.0176	-	-
2.9296	1705	0.0229	-	-
2.9381	1710	0.0166	-	-
2.9467	1715	0.0185	-	-
2.9553	1720	0.017	-	-
2.9639	1725	0.0109	-	-
2.9725	1730	0.0154	-	-
2.9811	1735	0.0226	-	-
2.9897	1740	0.0142	-	-
2.9983	1745	0.0257	-	-

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.1
• Transformers: 4.47.1
• PyTorch: 2.5.1+cu121
• Accelerate: 1.2.1
• Datasets: 2.14.4
• 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",
}