NilsML/fine_tuned_miniLM

SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. 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: sentence-transformers/all-MiniLM-L6-v2
• Maximum Sequence Length: 256 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': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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})
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    'How many different main types of diabetes are there?',
    'two',
    'a connection between relativistic quantum mechanics in the foldy - wouthuysen representation and the paraxial equations is established for a dirac particle in external fields. the paraxial form of the landau eigenfunction for a relativistic electron in a uniform magnetic field is determined. the obtained wave function contains the gouy phase and significantly approaches to the paraxial wave function for a free electron.',
]
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]

Evaluation

Metrics

Information Retrieval

• Dataset: sciq-eval
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.084
cosine_accuracy@3	0.192
cosine_accuracy@5	0.26
cosine_accuracy@10	0.367
cosine_precision@1	0.084
cosine_precision@3	0.064
cosine_precision@5	0.052
cosine_precision@10	0.0367
cosine_recall@1	0.084
cosine_recall@3	0.192
cosine_recall@5	0.26
cosine_recall@10	0.367
cosine_ndcg@10	0.2077
cosine_mrr@10	0.1589
cosine_map@100	0.1741

Training Details

Training Dataset

Unnamed Dataset

• Size: 46,716 training samples
• Columns: sentence_0, sentence_1, and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 7 tokens mean: 17.62 tokens max: 63 tokens	min: 3 tokens mean: 92.89 tokens max: 133 tokens	min: 0.0 mean: 0.23 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  Unlike plants, animal species rely almost exclusively on what type of reproduction?	0. 4, \ sim 4, and \ sim 300 \ mum are stronger than 10 ^ 5, 10 ^ 8, and 10 ^ 4 times those of the local interstellar radiation field ( isrf ). below these values, the chemical pumping is the dominant source of excitation of the j > 1 levels, even at high kinetic temperatures ( \ sim 1000 k ). the far - infrared emission lines of ch + observed in the orion bar and the ngc 7027 pdrs are consistent with the predictions of our excitation model assuming an incident far - ultraviolet ( fuv ) radiation field of \ sim 3 \ times 10	0.0
  What type of energy occurs by splitting the nuclei of radioactive uranium?	we study the potential of future electron - ion collider ( eic ) data to probe four - fermion operators in the standard model effective field theory ( smeft ). the ability to perform measurements with both polarized electron and proton beams at the eic provides a powerful tool that can disentangle the effects from different smeft operators. we compare the potential constraints from an eic with those obtained from drell - yan data at the large hadron collider. we show that eic data plays an important complementary role since it probes combinations of wilson coefficients not accessible through available drell - yan measurements.	0.0
  What element, which often forms polymers, has a unique ability to form covalent bonds with many other atoms?	some divergent series $ f $. the convergence sets on $ \ gamma : = \ { [ 1 : z : \ psi ( z ) ] : z \ in \ mathbb { c } \ } \ subset \ mathbb { c } ^ 2 \ subset \ mathbb { p } ^ 2 $, where $ \ psi $ is a transcendental entire holomorphic function, are also studied and we obtain that a subset on $ \ gamma $ is a convergence set in $ \ mathbb { p } ^ 2 $ if and only if it is a countable union of compact projectively convex sets, and	0.0

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• num_train_epochs: 1
• multi_dataset_batch_sampler: round_robin

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: 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}
• tp_size: 0
• 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
• 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

Epoch	Step	Training Loss	sciq-eval_cosine_ndcg@10
0.0685	100	-	0.1200
0.1370	200	-	0.1562
0.2055	300	-	0.1780
0.2740	400	-	0.1811
0.3425	500	3.1705	0.1909
0.4110	600	-	0.1904
0.4795	700	-	0.1955
0.5479	800	-	0.2031
0.6164	900	-	0.2014
0.6849	1000	2.9054	0.2002
0.7534	1100	-	0.2058
0.8219	1200	-	0.2083
0.8904	1300	-	0.2084
0.9589	1400	-	0.2076
1.0	1460	-	0.2077

Framework Versions

• Python: 3.12.8
• Sentence Transformers: 3.4.1
• Transformers: 4.51.3
• PyTorch: 2.5.1+cu124
• Accelerate: 1.3.0
• Datasets: 3.2.0
• 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",
}

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}
}