agentlans/distilbert-base-multilingual-cased-aligned

SentenceTransformer based on distilbert/distilbert-base-multilingual-cased

This is a sentence-transformers model finetuned from distilbert/distilbert-base-multilingual-cased. 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-multilingual-cased
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 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: 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("agentlans/distilbert-base-multilingual-cased-aligned")
# Run inference
sentences = [
    'Palm DOC Conduit for KPilot',
    'PalmDOC- conduit foar KPilot',
    'Man nepatinka gyventi kaime.',
]
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]

Training Details

Training Dataset

Unnamed Dataset

• Size: 867,042 training samples
• Columns: sentence_0 and sentence_1
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1
  type	string	string
  details	min: 3 tokens mean: 21.88 tokens max: 121 tokens	min: 3 tokens mean: 30.11 tokens max: 230 tokens

• Samples:

  sentence_0	sentence_1
  They need to be internationally recognized and supported.	Mereka harus diakui dan dibantu secara internasional.
  I ride with these kids once a week, every Tuesday.	Ik rijd met deze kinderen een keer per week, elke dinsdag.
  We still have some.	අපි ගාව තව ඒවා තියෙනවනේ.

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• num_train_epochs: 1
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 8
• per_device_eval_batch_size: 8
• 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
• num_train_epochs: 1
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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: False
• 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
• 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: round_robin

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0046	500	0.1996
0.0092	1000	0.087
0.0138	1500	0.0771
0.0185	2000	0.0646
0.0231	2500	0.0443
0.0277	3000	0.0526
0.0323	3500	0.05
0.0369	4000	0.0479
0.0415	4500	0.0477
0.0461	5000	0.0427
0.0507	5500	0.0343
0.0554	6000	0.0358
0.0600	6500	0.0452
0.0646	7000	0.0397
0.0692	7500	0.0289
0.0738	8000	0.0274
0.0784	8500	0.0364
0.0830	9000	0.0283
0.0877	9500	0.0295
0.0923	10000	0.0337
0.0969	10500	0.0303
0.1015	11000	0.0252
0.1061	11500	0.0241
0.1107	12000	0.0225
0.1153	12500	0.0263
0.1199	13000	0.0255
0.1246	13500	0.0311
0.1292	14000	0.0201
0.1338	14500	0.0209
0.1384	15000	0.0205
0.1430	15500	0.0242
0.1476	16000	0.0332
0.1522	16500	0.0346
0.1569	17000	0.0225
0.1615	17500	0.0245
0.1661	18000	0.0166
0.1707	18500	0.0196
0.1753	19000	0.0264
0.1799	19500	0.0212
0.1845	20000	0.0201
0.1891	20500	0.0238
0.1938	21000	0.0175
0.1984	21500	0.022
0.2030	22000	0.0201
0.2076	22500	0.0197
0.2122	23000	0.0137
0.2168	23500	0.017
0.2214	24000	0.031
0.2261	24500	0.0238
0.2307	25000	0.0194
0.2353	25500	0.024
0.2399	26000	0.022
0.2445	26500	0.0276
0.2491	27000	0.016
0.2537	27500	0.0203
0.2583	28000	0.0245
0.2630	28500	0.0161
0.2676	29000	0.0132
0.2722	29500	0.0142
0.2768	30000	0.0171
0.2814	30500	0.0207
0.2860	31000	0.0189
0.2906	31500	0.0169
0.2953	32000	0.0225
0.2999	32500	0.0224
0.3045	33000	0.0114
0.3091	33500	0.0213
0.3137	34000	0.0146
0.3183	34500	0.0154
0.3229	35000	0.0218
0.3275	35500	0.0096
0.3322	36000	0.0147
0.3368	36500	0.0186
0.3414	37000	0.0214
0.3460	37500	0.0231
0.3506	38000	0.0165
0.3552	38500	0.0157
0.3598	39000	0.0128
0.3645	39500	0.018
0.3691	40000	0.0183
0.3737	40500	0.0203
0.3783	41000	0.02
0.3829	41500	0.0165
0.3875	42000	0.0128
0.3921	42500	0.0106
0.3967	43000	0.0174
0.4014	43500	0.0168
0.4060	44000	0.0114
0.4106	44500	0.0158
0.4152	45000	0.0108
0.4198	45500	0.0141
0.4244	46000	0.0137
0.4290	46500	0.0137
0.4337	47000	0.0215
0.4383	47500	0.0123
0.4429	48000	0.0138
0.4475	48500	0.0152
0.4521	49000	0.0144
0.4567	49500	0.016
0.4613	50000	0.0132
0.4659	50500	0.0164
0.4706	51000	0.0155
0.4752	51500	0.0145
0.4798	52000	0.0173
0.4844	52500	0.02
0.4890	53000	0.0168
0.4936	53500	0.011
0.4982	54000	0.0116
0.5029	54500	0.009
0.5075	55000	0.0143
0.5121	55500	0.0111
0.5167	56000	0.0138
0.5213	56500	0.0104
0.5259	57000	0.0146
0.5305	57500	0.0116
0.5351	58000	0.0157
0.5398	58500	0.013
0.5444	59000	0.0144
0.5490	59500	0.0134
0.5536	60000	0.0114
0.5582	60500	0.0101
0.5628	61000	0.0164
0.5674	61500	0.0151
0.5721	62000	0.0138
0.5767	62500	0.0107
0.5813	63000	0.0102
0.5859	63500	0.0153
0.5905	64000	0.0103
0.5951	64500	0.0136
0.5997	65000	0.0107
0.6043	65500	0.0101
0.6090	66000	0.0101
0.6136	66500	0.0117
0.6182	67000	0.0113
0.6228	67500	0.0131
0.6274	68000	0.0068
0.6320	68500	0.0053
0.6366	69000	0.0113
0.6413	69500	0.0119
0.6459	70000	0.0094
0.6505	70500	0.0072
0.6551	71000	0.0171
0.6597	71500	0.0121
0.6643	72000	0.0134
0.6689	72500	0.0147
0.6735	73000	0.0075
0.6782	73500	0.0125
0.6828	74000	0.0064
0.6874	74500	0.0071
0.6920	75000	0.0073
0.6966	75500	0.0075
0.7012	76000	0.0097
0.7058	76500	0.01
0.7105	77000	0.0123
0.7151	77500	0.0093
0.7197	78000	0.0103
0.7243	78500	0.0179
0.7289	79000	0.0091
0.7335	79500	0.0121
0.7381	80000	0.0104
0.7428	80500	0.0083
0.7474	81000	0.0116
0.7520	81500	0.0084
0.7566	82000	0.0077
0.7612	82500	0.0081
0.7658	83000	0.0101
0.7704	83500	0.0093
0.7750	84000	0.0095
0.7797	84500	0.0107
0.7843	85000	0.0108
0.7889	85500	0.0095
0.7935	86000	0.0082
0.7981	86500	0.0103
0.8027	87000	0.0069
0.8073	87500	0.009
0.8120	88000	0.0081
0.8166	88500	0.0074
0.8212	89000	0.0069
0.8258	89500	0.0066
0.8304	90000	0.0065
0.8350	90500	0.0065
0.8396	91000	0.0088
0.8442	91500	0.008
0.8489	92000	0.0069
0.8535	92500	0.0095
0.8581	93000	0.0082
0.8627	93500	0.0068
0.8673	94000	0.006
0.8719	94500	0.0082
0.8765	95000	0.0121
0.8812	95500	0.0098
0.8858	96000	0.0083
0.8904	96500	0.008
0.8950	97000	0.0053
0.8996	97500	0.0102
0.9042	98000	0.0093
0.9088	98500	0.0042
0.9134	99000	0.0093
0.9181	99500	0.0138
0.9227	100000	0.0105
0.9273	100500	0.0079
0.9319	101000	0.0118
0.9365	101500	0.0072
0.9411	102000	0.0094
0.9457	102500	0.0108
0.9504	103000	0.0092
0.9550	103500	0.0062
0.9596	104000	0.0073
0.9642	104500	0.0089
0.9688	105000	0.0092
0.9734	105500	0.0076
0.9780	106000	0.0103
0.9826	106500	0.0064
0.9873	107000	0.0072
0.9919	107500	0.0052
0.9965	108000	0.0061

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.0
• Transformers: 4.46.3
• PyTorch: 2.5.1+cu124
• Accelerate: 1.1.1
• Datasets: 3.1.0
• Tokenizers: 0.20.3

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",
}

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}
}