rsl-ai/octen-embedding-0.6-finetuned

SentenceTransformer based on Octen/Octen-Embedding-0.6B

This is a sentence-transformers model finetuned from Octen/Octen-Embedding-0.6B on the csv dataset. 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: Octen/Octen-Embedding-0.6B
• Maximum Sequence Length: 64 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • csv

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': 64, 'do_lower_case': False, 'architecture': 'Qwen3Model'})
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': False, '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': True, '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("rsl-ai/octen-embedding-0.6-finetuned")
# Run inference
queries = [
    "ford transit ft 350 l td asm",
]
documents = [
    'ford transit ft 350 l td',
    'honda edix 2.0i',
    'bmw 3er 325i',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 1024] [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.7983, 0.0362, 0.0948]])

Evaluation

Metrics

Semantic Similarity

• Dataset: dev-positive-pairs
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	nan
spearman_cosine	nan

Training Details

Training Dataset

csv

• Dataset: csv
• Size: 19,662 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 4 tokens mean: 14.85 tokens max: 45 tokens	min: 6 tokens mean: 16.96 tokens max: 47 tokens

• Samples:

  anchor	positive
  citroen c25 1400 948	citroen c 25 1400 948
  volkswagen xl1	volkswagen xl1 0.8td
  skoda roomster 1.4 scout	skoda roomster 1.4i scout

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Evaluation Dataset

csv

• Dataset: csv
• Size: 1,035 evaluation samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 3 tokens mean: 15.63 tokens max: 42 tokens	min: 7 tokens mean: 17.45 tokens max: 43 tokens

• Samples:

  anchor	positive
  вазlada granta 1.6 luxe prestige 21927-a2-5yg	вазlada 2190 granta 1.6i club 21907-a1-v02
  toyota land cruiser 4.0dual vvt-i sport	toyota land cruiser 4.5i
  toyota land cruiser 3.0td 4wd	toyota land cruiser 3.4d

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 32
• learning_rate: 2e-05
• lr_scheduler_type: cosine
• bf16: True
• dataloader_num_workers: 2
• dataloader_prefetch_factor: 2
• gradient_checkpointing: True
• eval_on_start: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 32
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 2e-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: 3
• max_steps: -1
• lr_scheduler_type: cosine
• lr_scheduler_kwargs: None
• warmup_ratio: None
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• enable_jit_checkpoint: False
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• use_cpu: False
• seed: 42
• data_seed: None
• bf16: True
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: -1
• ddp_backend: None
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 2
• dataloader_prefetch_factor: 2
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• group_by_length: False
• length_column_name: length
• project: huggingface
• trackio_space_id: trackio
• 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
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• hub_revision: None
• gradient_checkpointing: True
• gradient_checkpointing_kwargs: None
• include_for_metrics: []
• eval_do_concat_batches: True
• auto_find_batch_size: False
• full_determinism: False
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• include_num_input_tokens_seen: no
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: True
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: True
• use_cache: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	dev-positive-pairs_spearman_cosine
0	0	-	0.1335	nan
0.0407	50	0.0100	-	-
0.0814	100	0.0187	-	-
0.1221	150	0.0160	-	-
0.1627	200	0.0155	-	-
0.2034	250	0.0141	-	-
0.2441	300	0.0133	-	-
0.2848	350	0.0136	-	-
0.3255	400	0.0134	-	-
0.3662	450	0.0083	-	-
0.4068	500	0.0310	-	-
0.4475	550	0.0086	-	-
0.4882	600	0.0197	-	-
0.5289	650	0.0174	-	-
0.5696	700	0.0086	-	-
0.6103	750	0.0108	-	-
0.6509	800	0.0275	-	-
0.6916	850	0.0072	-	-
0.7323	900	0.0080	-	-
0.7730	950	0.0220	-	-
0.8137	1000	0.0144	-	-
0.8544	1050	0.0118	-	-
0.8950	1100	0.0087	-	-
0.9357	1150	0.0075	-	-
0.9764	1200	0.0062	-	-
1.0	1229	-	0.0205	nan
1.0171	1250	0.0076	-	-
1.0578	1300	0.0092	-	-
1.0985	1350	0.0107	-	-
1.1391	1400	0.0059	-	-
1.1798	1450	0.0098	-	-
1.2205	1500	0.0037	-	-
1.2612	1550	0.0069	-	-
1.3019	1600	0.0097	-	-
1.3426	1650	0.0030	-	-
1.3832	1700	0.0106	-	-
1.4239	1750	0.0157	-	-
1.4646	1800	0.0084	-	-
1.5053	1850	0.0046	-	-
1.5460	1900	0.0103	-	-
1.5867	1950	0.0025	-	-
1.6273	2000	0.0080	-	-
1.6680	2050	0.0151	-	-
1.7087	2100	0.0099	-	-
1.7494	2150	0.0018	-	-
1.7901	2200	0.0072	-	-
1.8308	2250	0.0018	-	-
1.8714	2300	0.0096	-	-
1.9121	2350	0.0024	-	-
1.9528	2400	0.0056	-	-
1.9935	2450	0.0040	-	-
2.0	2458	-	0.0231	nan
2.0342	2500	0.0029	-	-
2.0749	2550	0.0108	-	-
2.1155	2600	0.0036	-	-
2.1562	2650	0.0080	-	-
2.1969	2700	0.0007	-	-
2.2376	2750	0.0027	-	-
2.2783	2800	0.0070	-	-
2.3190	2850	0.0048	-	-
2.3596	2900	0.0055	-	-
2.4003	2950	0.0083	-	-
2.4410	3000	0.0030	-	-
2.4817	3050	0.0131	-	-
2.5224	3100	0.0018	-	-
2.5631	3150	0.0041	-	-
2.6037	3200	0.0106	-	-
2.6444	3250	0.0117	-	-
2.6851	3300	0.0014	-	-
2.7258	3350	0.0059	-	-
2.7665	3400	0.0042	-	-
2.8072	3450	0.0122	-	-
2.8478	3500	0.0027	-	-
2.8885	3550	0.0118	-	-
2.9292	3600	0.0055	-	-
2.9699	3650	0.0147	-	-
3.0	3687	-	0.0225	nan

Framework Versions

• Python: 3.12.13
• Sentence Transformers: 5.3.0
• Transformers: 5.0.0
• PyTorch: 2.10.0+cu128
• Accelerate: 1.13.0
• Datasets: 4.0.0
• Tokenizers: 0.22.2

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{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}