Mihaiii/test33

SentenceTransformer based on Mihaiii/Venusaur

This is a sentence-transformers model finetuned from Mihaiii/Venusaur on the Mihaiii/qa-assistant-2 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: Mihaiii/Venusaur
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 tokens
• Similarity Function: Cosine Similarity
• Training Dataset:
  • Mihaiii/qa-assistant-2
• Language: en

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': 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})
)

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 = [
    'What are the social structures of ants?',
    'The social hierarchy of ants is a complex system that ensures the survival and efficiency of the colony.',
    'In a parliamentary system, the executive branch derives its legitimacy from and is accountable to the legislature; the executive and legislative branches are thus interconnected.',
]
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

Semantic Similarity

• Dataset: sts-dev
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8261
spearman_cosine	0.8277
pearson_manhattan	0.82
spearman_manhattan	0.8226
pearson_euclidean	0.8215
spearman_euclidean	0.8237
pearson_dot	0.8037
spearman_dot	0.8082
pearson_max	0.8261
spearman_max	0.8277

Training Details

Training Dataset

Mihaiii/qa-assistant-2

• Dataset: Mihaiii/qa-assistant-2 at 9650e69
• Size: 16,011 training samples
• Columns: question, answer, and score
• Approximate statistics based on the first 1000 samples:

  	question	answer	score
  type	string	string	float
  details	min: 6 tokens mean: 12.73 tokens max: 27 tokens	min: 10 tokens mean: 22.42 tokens max: 65 tokens	min: 0.02 mean: 0.53 max: 1.0

• Samples:

  question	answer	score
  Can you describe the process of robot path planning?	Robots can be programmed to perform a variety of tasks, from simple repetitive actions to complex decision-making processes.	0.27999999999999997
  Can humans live on Mars?	Mars is the fourth planet from the Sun and is often called the Red Planet due to its reddish appearance.	0.16
  What are the key elements of composition in abstract art?	The history of abstract art dates back to the early 20th century, with pioneers like Wassily Kandinsky and Piet Mondrian.	0.36

• Loss: CosineSimilarityLoss with these parameters:

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

Evaluation Dataset

Mihaiii/qa-assistant-2

• Dataset: Mihaiii/qa-assistant-2 at 9650e69
• Size: 3,879 evaluation samples
• Columns: question, answer, and score
• Approximate statistics based on the first 1000 samples:

  	question	answer	score
  type	string	string	float
  details	min: 7 tokens mean: 12.71 tokens max: 31 tokens	min: 10 tokens mean: 22.63 tokens max: 51 tokens	min: 0.02 mean: 0.53 max: 1.0

• Samples:

  question	answer	score
  What is the concept of social stratification?	The study of social stratification involves examining the inequalities and divisions within a society.	0.6799999999999999
  How does J.K. Rowling develop the character of Hermione Granger throughout the 'Harry Potter' series?	The 'Harry Potter' series consists of seven books, starting with 'Harry Potter and the Philosopher's Stone' and ending with 'Harry Potter and the Deathly Hallows'.	0.22000000000000003
  What is the parliamentary system and how does it function?	In a parliamentary system, the government can be dissolved by a vote of no confidence, which can lead to new elections.	0.6799999999999999

• 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: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 4
• warmup_ratio: 0.1

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_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: 1.0
• num_train_epochs: 4
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• 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
• 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
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	sts-dev_spearman_cosine
0.0999	100	0.0593	0.0540	0.5848
0.1998	200	0.05	0.0463	0.6618
0.2997	300	0.044	0.0418	0.7102
0.3996	400	0.0413	0.0385	0.7390
0.4995	500	0.0377	0.0349	0.7707
0.5994	600	0.034	0.0333	0.7770
0.6993	700	0.0344	0.0321	0.7879
0.7992	800	0.0324	0.0311	0.7927
0.8991	900	0.0334	0.0302	0.8005
0.9990	1000	0.0304	0.0305	0.8023
1.0989	1100	0.0261	0.0306	0.8072
1.1988	1200	0.0267	0.0292	0.8104
1.2987	1300	0.0244	0.0287	0.8110
1.3986	1400	0.0272	0.0294	0.8098
1.4985	1500	0.0241	0.0281	0.8135
1.5984	1600	0.0253	0.0282	0.8143
1.6983	1700	0.0245	0.0276	0.8169
1.7982	1800	0.025	0.0274	0.8182
1.8981	1900	0.0236	0.0273	0.8193
1.9980	2000	0.0236	0.0269	0.8218
2.0979	2100	0.0215	0.0278	0.8213
2.1978	2200	0.0216	0.0269	0.8226
2.2977	2300	0.0205	0.0276	0.8207
2.3976	2400	0.0181	0.0273	0.8202
2.4975	2500	0.0197	0.0267	0.8228
2.5974	2600	0.02	0.0267	0.8238
2.6973	2700	0.0203	0.0263	0.8258
2.7972	2800	0.0184	0.0263	0.8264
2.8971	2900	0.0201	0.0269	0.8243
2.9970	3000	0.0196	0.0263	0.8251
3.0969	3100	0.0168	0.0264	0.8250
3.1968	3200	0.0176	0.0263	0.8267
3.2967	3300	0.0168	0.0263	0.8270
3.3966	3400	0.017	0.0260	0.8277
3.4965	3500	0.0164	0.0261	0.8273
3.5964	3600	0.0172	0.0259	0.8280
3.6963	3700	0.0168	0.0260	0.8274
3.7962	3800	0.0176	0.0262	0.8279
3.8961	3900	0.0182	0.0261	0.8278
3.9960	4000	0.0174	0.0260	0.8277

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.0.1+cu118
• Accelerate: 0.31.0
• Datasets: 2.20.0
• Tokenizers: 0.19.1

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",
}