zacCMU/miniLM2-ENG

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

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2 on the rag_finetuning_for_engineering dataset. 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
• Training Dataset:
  • rag_finetuning_for_engineering

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, 'architecture': '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("zacCMU/miniLM2-ENG")
# Run inference
sentences = [
    '10.7 Information to be provided to the FIA and Competitors \na) In order that an FIA observer may be appointed, Competitors must inform the FIA and all \nother Competitors of any planned TPC, PE or DE at least 72 hours before it is due to \ncommence, and the following information must be provided: \ni) The precise specification of the car(s) to be used. ii) The name(s) of the driver(s). iii) The type of activity.',
    'Competitors must notify the FIA and other teams at least 72 hours in advance of any planned technical testing, physical evaluations, or development exercises, providing detailed information about the cars, drivers, and nature of the activity.',
    "The aerodynamic design of a Formula 1 car's rear wing is crucial in determining its overall downforce and drag characteristics, requiring a delicate balance between speed and stability.",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000, -0.6238,  0.9745],
#         [-0.6238,  1.0000, -0.6029],
#         [ 0.9745, -0.6029,  1.0000]])

Training Details

Training Dataset

rag_finetuning_for_engineering

• Dataset: rag_finetuning_for_engineering at bddb325
• Size: 637 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 637 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 9 tokens mean: 112.88 tokens max: 256 tokens	min: 18 tokens mean: 60.28 tokens max: 176 tokens	min: 15 tokens mean: 35.03 tokens max: 57 tokens

• Samples:

  anchor	positive	negative
  A penalty in accordance with Article 54.3d) will be imposed on any driver who fails to start the race from the pit lane. If any driver needs assistance after the fifteen (15) second signal, he must raise his arm and, when the remainder of the cars able to do so have left the pit lane, marshals will be instructed to push the car into the inner lane. In this case, marshals with yellow flags will stand beside any car concerned to warn drivers behind.	A driver who fails to start the race from the pit lane will incur a penalty. If a driver requires assistance after the 15-second signal, they must signal for help and marshals will then guide their car into the inner lane, warning other drivers with yellow flags.	The aerodynamic design of modern Formula 1 cars requires a delicate balance between downforce and drag to achieve optimal speed on the track.
  If a driver wishes to leave his car before it is weighed , he must ask the Technical Delegate to weigh him in order that this weight may be added to that of the car. e) If a car stops on the circuit during the qualifying session or the sprint qualifying session and the driver leaves the car, he must go to the FIA garage immediately on his return to the pit lane in order for his weight to be established. 35.2 After the sprint session or the race any classified car may be weighed.	To avoid penalties, a driver must ensure their weight is accurately recorded before leaving their car, either by having the Technical Delegate weigh them or by being weighed in the FIA garage after returning to the pit lane. This process is crucial during qualifying sessions, sprint qualifying sessions, and after the sprint session or the race.	The aerodynamic design of a Formula 1 car's rear wing plays a crucial role in generating downforce, but its impact on the overall handling and stability of the vehicle is often overlooked by teams in their pursuit of speed.
  d) When leaving the pits a driver may overtake, or be overtaken by, another car on the track before he reaches the second safety car line. e) When the safety car is returning to the pits it may be overtaken by cars on the track once it has reached the first safety car line. f) Whilst in the pit entry road, pit lane or pit exit road a driver may overtake another car which is also in one of these three areas.	When exiting the pits, a driver is allowed to overtake or be overtaken by another car on the track before reaching the second safety car line. Additionally, the safety car can be overtaken by cars on the track once it has reached the first safety car line, and drivers can also overtake each other while in the pit entry road, pit lane, or pit exit road.	The aerodynamic design of modern Formula 1 cars relies heavily on complex computational fluid dynamics simulations to optimize their downforce and drag characteristics.

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
      "triplet_margin": 5
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 16
• learning_rate: 1e-05
• num_train_epochs: 4

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 16
• 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: 1e-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: 4
• 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
• 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}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• 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
• 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
• hub_revision: None
• 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: no
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: True
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.25	10	5.5019
0.5	20	5.2724
0.75	30	5.1275
1.0	40	4.999
1.25	50	4.8488
1.5	60	4.7919
1.75	70	4.6734
2.0	80	4.4696
2.25	90	4.4078
2.5	100	4.2232
2.75	110	4.1736
3.0	120	4.0837
3.25	130	4.0113
3.5	140	4.0376
3.75	150	3.9134
4.0	160	3.9853

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.1.2
• Transformers: 4.57.1
• PyTorch: 2.9.0+cu126
• Accelerate: 1.11.0
• Datasets: 4.0.0
• Tokenizers: 0.22.1

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}