aisuko/training

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 csv 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:
  • 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': 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("training")
# Run inference
sentences = [
    '<think>\nLet’s think through this step by step\nn = 4\nsp = 75\nt = 36\nspt = 36 / 4 = 9\nop = 75 + 9 = 84\n</think>\n\\boxed{84}',
    '<think>\nLet’s think through this step by step\nn = 4\nsp = 75\nt = 36\nspt = t / n = 36 / 4 = 9\nop = sp + spt = 75 + 9 = 84\n</think>\n\\boxed{84}',
    "<think>\nLet’s think through this step by step\nLet B be Benedict's house size\nK = 10000 sq ft\nK = 4B + 600\n10000 = 4B + 600\n4B = 9400\nB = 2350 sq ft\n</think>\n\\boxed{2350}",
]
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

• Datasets: sts-dev and sts-test
• Evaluated with EmbeddingSimilarityEvaluator

Metric	sts-dev	sts-test
pearson_cosine	0.8793	0.8793
spearman_cosine	0.8765	0.8765

Training Details

Training Dataset

csv

• Dataset: csv
• Size: 4,984 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 34 tokens mean: 66.41 tokens max: 256 tokens	min: 36 tokens mean: 67.03 tokens max: 256 tokens

• Samples:

  anchor	positive
  Let’s think through this step by step ht_hare = 14 inches ht_camel = 24 × 14 = 336 inches ht_camel_ft = 336 / 12 = 28 feet \boxed{28}	Let’s think through this step by step ht_hare = 14 inches ht_camel = 24 * 14 = 336 inches ht_camel_ft = 336 / 12 = 28 feet \boxed{28}
  Let’s think through this step by step cpb = 5 sp = 90 pmpb = 2 tp = 5 × 5 = 25 tpmp = 5 × 2 = 10 profit = 90 - 25 - 10 = 55 \boxed{55}	Let’s think through this step by step cp = 5 × 5 = 25 sp = 90 p = sp - cp - (5 × 2) = 90 - 25 - 10 = 55 \boxed{55}
  Let’s think through this step by step Time to iron a blouse = 15 minutes Time to iron a dress = 20 minutes Time spent on blouses = 2 hours = 120 minutes Time spent on dresses = 3 hours = 180 minutes Number of blouses ironed = 120 / 15 = 8 Number of dresses ironed = 180 / 20 = 9 Total pieces of clothes ironed = 8 + 9 = 17 \boxed{17}	Let’s think through this step by step Time to iron 1 blouse = 15 minutes Time to iron 1 dress = 20 minutes Time spent on blouses = 2 hours = 120 minutes Time spent on dresses = 3 hours = 180 minutes Number of blouses ironed = 120 / 15 = 8 Number of dresses ironed = 180 / 20 = 9 Total pieces of clothes ironed = 8 + 9 = 17 \boxed{17}

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

csv

• Dataset: csv
• Size: 4,984 evaluation samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 33 tokens mean: 66.68 tokens max: 164 tokens	min: 33 tokens mean: 66.71 tokens max: 161 tokens

• Samples:

  anchor	positive
  Let’s think through this step by step sg = 36 sc = 1/4 fl = 1/3 sg_left = sg - (sg × sc) - (sg × (1 - sc) × fl) sg_left = 36 - (36 × 1/4) - (36 × (1 - 1/4) × 1/3) sg_left = 36 - 9 - 6 sg_left = 21 \boxed{21}	Let’s think through this step by step sg = 36 sc = 1/4 fl = 1/3 sg_left = sg - (sg × sc) - (sg × (1 - sc) × fl) sg_left = 36 - (36 × 1/4) - (36 × (1 - 1/4) × 1/3) sg_left = 36 - 9 - 6 sg_left = 21 \boxed{21}
  Let’s think through this step by step d1 = 125 miles d2 = 223 miles d3 = 493 - (125 + 223) d3 = 145 miles \boxed{145}	Let’s think through this step by step d1 = 125 miles d2 = 223 miles d3 = 493 - (125 + 223) d3 = 145 miles \boxed{145}
  Let’s think through this step by step Total workdays = 2 weeks × 5 days/week = 10 days Paid vacation days = 6 days Unpaid vacation days = 10 - 6 = 4 days Total pay = 15 × 8 = $120/day Missed pay = 4 × 120 = $480 \boxed{480}	Let’s think through this step by step Total workdays = 2 weeks × 5 days/week = 10 days Paid vacation days = 6 days Unpaid vacation days = 10 - 6 = 4 days Total pay = 15 × 8 = $120/day Missed pay = 4 × 120 = $480 \boxed{480}

• 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: 16
• per_device_eval_batch_size: 16
• learning_rate: 2e-05
• num_train_epochs: 20
• warmup_ratio: 0.1
• fp16: True
• batch_sampler: no_duplicates

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
• 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: 20
• 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: True
• 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: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	sts-dev_spearman_cosine	sts-test_spearman_cosine
-1	-1	-	-	0.8671	-
0.3953	100	0.0422	0.0031	0.8701	-
0.7905	200	0.0105	0.0017	0.8727	-
1.1858	300	0.0041	0.0016	0.8728	-
1.5810	400	0.0016	0.0011	0.8730	-
1.9763	500	0.0039	0.0021	0.8731	-
2.3715	600	0.0014	0.0020	0.8741	-
2.7668	700	0.0014	0.0017	0.8744	-
3.1621	800	0.0019	0.0009	0.8742	-
3.5573	900	0.0012	0.0011	0.8754	-
3.9526	1000	0.0016	0.0015	0.8760	-
4.3478	1100	0.0021	0.0011	0.8763	-
4.7431	1200	0.0006	0.0009	0.8753	-
5.1383	1300	0.0004	0.0009	0.8753	-
5.5336	1400	0.0008	0.0008	0.8751	-
5.9289	1500	0.0004	0.0004	0.8743	-
6.3241	1600	0.0009	0.0008	0.8758	-
6.7194	1700	0.0005	0.0009	0.8747	-
7.1146	1800	0.0004	0.0006	0.8742	-
7.5099	1900	0.0003	0.0010	0.8748	-
7.9051	2000	0.0006	0.0008	0.8742	-
8.3004	2100	0.0005	0.0007	0.8744	-
8.6957	2200	0.0003	0.0006	0.8748	-
9.0909	2300	0.0005	0.0012	0.8749	-
9.4862	2400	0.0007	0.0006	0.8762	-
9.8814	2500	0.0003	0.0009	0.8762	-
10.2767	2600	0.0004	0.0007	0.8759	-
10.6719	2700	0.0005	0.0005	0.8760	-
11.0672	2800	0.0005	0.0007	0.8754	-
11.4625	2900	0.0002	0.0008	0.8749	-
11.8577	3000	0.0002	0.0007	0.8749	-
12.2530	3100	0.0003	0.0007	0.8752	-
12.6482	3200	0.0004	0.0008	0.8760	-
13.0435	3300	0.0002	0.0008	0.8767	-
13.4387	3400	0.0002	0.0007	0.8763	-
13.8340	3500	0.0002	0.0007	0.8763	-
14.2292	3600	0.0001	0.0007	0.8764	-
14.6245	3700	0.0003	0.0006	0.8765	-
15.0198	3800	0.0002	0.0005	0.8757	-
15.4150	3900	0.0002	0.0004	0.8760	-
15.8103	4000	0.0002	0.0005	0.8765	-
16.2055	4100	0.0002	0.0005	0.8757	-
16.6008	4200	0.0002	0.0006	0.8758	-
16.9960	4300	0.0002	0.0006	0.8758	-
17.3913	4400	0.0001	0.0005	0.8761	-
17.7866	4500	0.0002	0.0005	0.8765	-
18.1818	4600	0.0001	0.0005	0.8767	-
18.5771	4700	0.0004	0.0004	0.8765	-
18.9723	4800	0.0002	0.0004	0.8765	-
19.3676	4900	0.0001	0.0004	0.8765	-
19.7628	5000	0.0001	0.0004	0.8765	-
-1	-1	-	-	-	0.8765

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 3.4.1
• Transformers: 4.51.1
• PyTorch: 2.5.1+cu124
• Accelerate: 1.3.0
• Datasets: 3.5.0
• 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}
}