Jimmy-Ooi/TTM_800_8_8_0.0001_AdamW

SentenceTransformer based on google-bert/bert-base-cased

This is a sentence-transformers model finetuned from google-bert/bert-base-cased on the csv dataset. It maps sentences & paragraphs to a 768-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: google-bert/bert-base-cased
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 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': 512, 'do_lower_case': False, 'architecture': 'BertModel'})
  (1): Pooling({'word_embedding_dimension': 768, '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("Jimmy-Ooi/TTM_800_8_8_0.0001_AdamW")
# Run inference
sentences = [
    'O=C(/C=C/c1ccc(O)c(O)c1)NC(Cc1ccccc1)C(=O)NO',
    'Cc1cccc(C(=O)Nc2cccc(C(=O)/C=C/c3ccc4c(c3)c3ccccc3n4C)c2)c1',
    'COc1ccc(CCc2ccc(O)cc2O)cc1',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000, -0.1056, -0.2395],
#         [-0.1056,  1.0000,  0.9876],
#         [-0.2395,  0.9876,  1.0000]])

Training Details

Training Dataset

csv

• Dataset: csv
• Size: 120,059 training samples
• Columns: premise, hypothesis, and label
• Approximate statistics based on the first 1000 samples:

  	premise	hypothesis	label
  type	string	string	int
  details	min: 11 tokens mean: 39.64 tokens max: 106 tokens	min: 11 tokens mean: 38.71 tokens max: 106 tokens	0: ~52.50% 2: ~47.50%

• Samples:

  premise	hypothesis	label
  C/C(=N\NC(N)=S)c1cccc(N)c1	COc1ccc(C(=O)OCc2cc(=O)c(O)co2)c(OC)c1	2
  CC(=O)N1N=C(c2ccc(O)cc2)CC1c1ccc(O)cc1	O=C(OCc1ccc(O)cc1)c1cc(O)ccc1O	0
  OC[C@H]1OC@HC@HC@@H[C@@H]1O	COc1cc(C=O)ccc1OC(=O)CN1CCN(C)CC1	0

• Loss: SoftmaxLoss

Evaluation Dataset

csv

• Dataset: csv
• Size: 21,187 evaluation samples
• Columns: premise, hypothesis, and label
• Approximate statistics based on the first 1000 samples:

  	premise	hypothesis	label
  type	string	string	int
  details	min: 11 tokens mean: 39.6 tokens max: 106 tokens	min: 12 tokens mean: 38.62 tokens max: 106 tokens	0: ~47.70% 2: ~52.30%

• Samples:

  premise	hypothesis	label
  COc1cc(C2CC(c3ccccc3O)=NN2C(C)=O)ccc1O	Cn1c2ccccc2c2cc(/C=C/C(=O)c3cccc(N)c3)ccc21	0
  CC(=S)Nc1ccccc1	O=C(NO)Nc1ccc(O)cc1	2
  C/C(=N\NC(N)=S)c1cccc(NC(=O)C(C)(C)C)c1	OC[C@H]1OC@@HC@HC@@H[C@@H]1O	0

• Loss: SoftmaxLoss

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 64
• num_train_epochs: 8
• warmup_steps: 100
• optim: adamw_torch
• weight_decay: 0.0001
• fp16: True
• per_device_eval_batch_size: 64

All Hyperparameters

Click to expand

• per_device_train_batch_size: 64
• num_train_epochs: 8
• max_steps: -1
• learning_rate: 5e-05
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_steps: 100
• optim: adamw_torch
• optim_args: None
• weight_decay: 0.0001
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• optim_target_modules: None
• gradient_accumulation_steps: 1
• average_tokens_across_devices: True
• max_grad_norm: 1.0
• label_smoothing_factor: 0.0
• bf16: False
• fp16: True
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• use_liger_kernel: False
• liger_kernel_config: None
• use_cache: False
• neftune_noise_alpha: None
• torch_empty_cache_steps: None
• auto_find_batch_size: False
• log_on_each_node: True
• logging_nan_inf_filter: True
• include_num_input_tokens_seen: no
• log_level: passive
• log_level_replica: warning
• disable_tqdm: False
• project: huggingface
• trackio_space_id: trackio
• eval_strategy: no
• per_device_eval_batch_size: 64
• prediction_loss_only: True
• eval_on_start: False
• eval_do_concat_batches: True
• eval_use_gather_object: False
• eval_accumulation_steps: None
• include_for_metrics: []
• batch_eval_metrics: False
• save_only_model: False
• save_on_each_node: False
• enable_jit_checkpoint: False
• push_to_hub: False
• hub_private_repo: None
• hub_model_id: None
• hub_strategy: every_save
• hub_always_push: False
• hub_revision: None
• load_best_model_at_end: False
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• data_seed: None
• use_cpu: 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
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• remove_unused_columns: True
• label_names: None
• train_sampling_strategy: random
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• ddp_backend: None
• ddp_timeout: 1800
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• deepspeed: None
• debug: []
• skip_memory_metrics: True
• do_predict: False
• resume_from_checkpoint: None
• warmup_ratio: None
• local_rank: -1
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0533	100	0.9256
0.1066	200	0.6870
0.1599	300	0.6471
0.2132	400	0.6166
0.2665	500	0.5977
0.3198	600	0.5855
0.3731	700	0.5860
0.4264	800	0.5744
0.4797	900	0.5765
0.5330	1000	0.5681
0.5864	1100	0.5625
0.6397	1200	0.5667
0.6930	1300	0.5581
0.7463	1400	0.5521
0.7996	1500	0.5476
0.8529	1600	0.5482
0.9062	1700	0.5588
0.9595	1800	0.5476
1.0128	1900	0.5382
1.0661	2000	0.5425
1.1194	2100	0.5534
1.1727	2200	0.5509
1.2260	2300	0.5356
1.2793	2400	0.5399
1.3326	2500	0.5328
1.3859	2600	0.5358
1.4392	2700	0.5393
1.4925	2800	0.5331
1.5458	2900	0.5202
1.5991	3000	0.5282
1.6525	3100	0.5331
1.7058	3200	0.5326
1.7591	3300	0.5331
1.8124	3400	0.5303
1.8657	3500	0.5285
1.9190	3600	0.5257
1.9723	3700	0.5358
2.0256	3800	0.5266
2.0789	3900	0.5316
2.1322	4000	0.5226
2.1855	4100	0.5271
2.2388	4200	0.5119
2.2921	4300	0.5294
2.3454	4400	0.5168
2.3987	4500	0.5218
2.4520	4600	0.5155
2.5053	4700	0.5239
2.5586	4800	0.5249
2.6119	4900	0.5222
2.6652	5000	0.5144
2.7186	5100	0.5211
2.7719	5200	0.5239
2.8252	5300	0.5101
2.8785	5400	0.5166
2.9318	5500	0.5235
2.9851	5600	0.5154
3.0384	5700	0.5170
3.0917	5800	0.5159
3.1450	5900	0.5190
3.1983	6000	0.5166
3.2516	6100	0.5116
3.3049	6200	0.5117
3.3582	6300	0.5038
3.4115	6400	0.5107
3.4648	6500	0.5143
3.5181	6600	0.5168
3.5714	6700	0.5002
3.6247	6800	0.5083
3.6780	6900	0.5130
3.7313	7000	0.5091
3.7846	7100	0.5164
3.8380	7200	0.5130
3.8913	7300	0.5101
3.9446	7400	0.4955
3.9979	7500	0.5065
4.0512	7600	0.5111
4.1045	7700	0.5077
4.1578	7800	0.5178
4.2111	7900	0.5068
4.2644	8000	0.5005
4.3177	8100	0.5010
4.3710	8200	0.5043
4.4243	8300	0.4964
4.4776	8400	0.5016
4.5309	8500	0.5069
4.5842	8600	0.5082
4.6375	8700	0.5083
4.6908	8800	0.5071
4.7441	8900	0.5082
4.7974	9000	0.5117
4.8507	9100	0.5023
4.9041	9200	0.4969
4.9574	9300	0.5069
5.0107	9400	0.5061
5.0640	9500	0.4983
5.1173	9600	0.5004
5.1706	9700	0.5047
5.2239	9800	0.4958
5.2772	9900	0.5076
5.3305	10000	0.5048
5.3838	10100	0.5018
5.4371	10200	0.5090
5.4904	10300	0.5019
5.5437	10400	0.5001
5.5970	10500	0.4958
5.6503	10600	0.5004
5.7036	10700	0.5025
5.7569	10800	0.5004
5.8102	10900	0.4970
5.8635	11000	0.5054
5.9168	11100	0.5062
5.9701	11200	0.4970
6.0235	11300	0.5052
6.0768	11400	0.5016
6.1301	11500	0.4965
6.1834	11600	0.4988
6.2367	11700	0.5051
6.2900	11800	0.4899
6.3433	11900	0.4931
6.3966	12000	0.5085
6.4499	12100	0.5011
6.5032	12200	0.4992
6.5565	12300	0.5007
6.6098	12400	0.5081
6.6631	12500	0.4911
6.7164	12600	0.4992
6.7697	12700	0.5037
6.8230	12800	0.4926
6.8763	12900	0.4957
6.9296	13000	0.4964
6.9829	13100	0.5046
7.0362	13200	0.5048
7.0896	13300	0.4954
7.1429	13400	0.5026
7.1962	13500	0.4963
7.2495	13600	0.4908
7.3028	13700	0.4945
7.3561	13800	0.4999
7.4094	13900	0.4935
7.4627	14000	0.4909
7.5160	14100	0.4917
7.5693	14200	0.4948
7.6226	14300	0.4942
7.6759	14400	0.5051
7.7292	14500	0.5062
7.7825	14600	0.4935
7.8358	14700	0.4935
7.8891	14800	0.5019
7.9424	14900	0.5043
7.9957	15000	0.4964

Framework Versions

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

Citation

BibTeX

Sentence Transformers and SoftmaxLoss

@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",
}