HassanCS/chemBERTa-tuned-on-ClinTox-using-triplet-loss

SentenceTransformer based on DeepChem/ChemBERTa-77M-MLM

This is a sentence-transformers model finetuned from DeepChem/ChemBERTa-77M-MLM. 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: DeepChem/ChemBERTa-77M-MLM
• Maximum Sequence Length: 512 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': 512, 'do_lower_case': False}) with Transformer model: RobertaModel 
  (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("HassanCS/chemBERTa-tuned-on-ClinTox-using-triplet-loss")
# Run inference
sentences = [
    'CON=C(C(=O)NC1C(=O)N2C(C(=O)[O-])=C(C[N+]3(C)CCCC3)CSC12)c1csc(N)n1',
    'CC1CNc2c(cccc2S(=O)(=O)NC(CCC[NH+]=C(N)N)C(=O)N2CCC(C)CC2C(=O)[O-])C1',
    'CC(C)C1(C(=O)NC2CC(=O)OC2(O)CF)CC(c2nccc3ccccc23)=NO1',
]
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

Triplet

• Dataset: all-dev
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.7135

Training Details

Training Dataset

Unnamed Dataset

• Size: 118,400 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 20 tokens mean: 33.0 tokens max: 60 tokens	min: 3 tokens mean: 47.34 tokens max: 212 tokens	min: 3 tokens mean: 53.88 tokens max: 212 tokens

• Samples:

  anchor	positive	negative
  CC(C)CC(NC(=O)CNC(=O)c1cc(Cl)ccc1Cl)B(O)O	CC(=O)OC1CCC2(C)C(=CCC3C2CCC2(C)C(c4cccnc4)=CCC32)C1	CCOC(=O)c1ncn2c1CN(C)C(=O)c1cc(F)ccc1-2
  CC(C)CC(NC(=O)CNC(=O)c1cc(Cl)ccc1Cl)B(O)O	COc1ccc(C(CN(C)C)C2(O)CCCCC2)cc1	C[NH2+]C1(C)C2CCC(C2)C1(C)C
  CC(C)CC(NC(=O)CNC(=O)c1cc(Cl)ccc1Cl)B(O)O	CNC(=O)c1cc(Oc2ccc(NC(=O)Nc3ccc(Cl)c(C(F)(F)F)c3)cc2)ccn1.Cc1ccc(S(=O)(=O)O)cc1	Nc1ncnc2c1ncn2C1OC(CO)C(O)C1O

• Loss: TripletLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 1,480 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 18 tokens mean: 54.07 tokens max: 169 tokens	min: 18 tokens mean: 60.4 tokens max: 244 tokens	min: 30 tokens mean: 71.25 tokens max: 141 tokens

• Samples:

  anchor	positive	negative
  CC(C)OC(=O)CCCC=CCC1C(O)CC(O)C1C=CC(O)COc1cccc(C(F)(F)F)c1	CC12CCCCCC(Cc3ccc(O)cc31)C2[NH3+]	CC(C)C(CN1CCC(C)(c2cccc(O)c2)C(C)C1)NC(=O)C1Cc2ccc(O)cc2CN1
  CC(C)OC(=O)CCCC=CCC1C(O)CC(O)C1C=CC(O)COc1cccc(C(F)(F)F)c1	COc1cc2c(cc1OC)C1CC(=O)C(CC(C)C)C[NH+]1CC2	CC(C)C(CN1CCC(C)(c2cccc(O)c2)C(C)C1)NC(=O)C1Cc2ccc(O)cc2CN1
  CC(C)OC(=O)CCCC=CCC1C(O)CC(O)C1C=CC(O)COc1cccc(C(F)(F)F)c1	CNH+CCC=C1c2ccccc2COc2ccc(CC(=O)[O-])cc21	CC(C)C1(C(=O)NC2CC(=O)OC2(O)CF)CC(c2nccc3ccccc23)=NO1

• Loss: TripletLoss with these parameters:

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

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: 5
• 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: 5
• 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}
• 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
• 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
• 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	all-dev_cosine_accuracy
0.0676	500	5.0821	5.1737	0.4047
0.1351	1000	4.9869	5.1766	0.4230
0.2027	1500	4.5562	4.9102	0.5345
0.2703	2000	3.2364	4.3712	0.6534
0.3378	2500	2.0738	4.0704	0.6736
0.4054	3000	1.4239	4.0200	0.6635
0.4730	3500	1.1578	3.7202	0.6791
0.5405	4000	0.9669	3.7197	0.6831
0.6081	4500	0.714	3.8818	0.6547
0.6757	5000	0.5359	4.0987	0.6243
0.7432	5500	0.5663	3.8127	0.6500
0.8108	6000	0.4827	3.8346	0.6676
0.8784	6500	0.4758	3.8333	0.6507
0.9459	7000	0.4759	3.6872	0.6912
1.0135	7500	0.4651	3.7229	0.6831
1.0811	8000	0.4739	3.8041	0.6662
1.1486	8500	0.4458	3.8235	0.6703
1.2162	9000	0.4189	3.7957	0.6716
1.2838	9500	0.4504	3.7422	0.6784
1.3514	10000	0.413	3.7588	0.6770
1.4189	10500	0.3808	3.9750	0.6615
1.4865	11000	0.3853	3.7417	0.6953
1.5541	11500	0.379	3.7319	0.6993
1.6216	12000	0.429	3.5620	0.7209
1.6892	12500	0.3735	3.6900	0.7020
1.7568	13000	0.3908	3.8182	0.6932
1.8243	13500	0.3848	3.7228	0.7101
1.8919	14000	0.3777	3.6604	0.7149
1.9595	14500	0.3912	3.7849	0.6946
2.0269	15000	0.3282	3.8607	0.7014
2.0945	15500	0.3324	3.8573	0.6953
2.1620	16000	0.3852	3.9420	0.7000
2.2296	16500	0.3633	3.7928	0.7189
2.2972	17000	0.3493	3.8217	0.7216
2.3647	17500	0.3554	3.8546	0.6993
2.4323	18000	0.3363	3.7764	0.6993
2.4999	18500	0.377	3.8224	0.6959
2.5674	19000	0.3569	3.8376	0.7155
2.635	19500	0.3414	4.0017	0.7034
2.7026	20000	0.3567	3.7405	0.7135
2.7701	20500	0.3524	3.9446	0.7189
2.8377	21000	0.3347	3.8140	0.7169
2.9053	21500	0.3458	4.0700	0.7088
2.9728	22000	0.3632	3.7930	0.7081
3.0404	22500	0.3496	3.9884	0.7236
3.1080	23000	0.3426	3.7102	0.7155
3.1755	23500	0.3579	3.9201	0.7135
3.2431	24000	0.3553	4.2237	0.7270
3.3107	24500	0.345	3.8090	0.7189
3.3782	25000	0.3475	3.7802	0.7284
3.4458	25500	0.3326	3.7549	0.7250
3.5134	26000	0.3228	3.6717	0.7216
3.5809	26500	0.3311	3.8241	0.7155
3.6485	27000	0.3215	3.8151	0.7142
3.7161	27500	0.3534	3.8639	0.7149
3.7836	28000	0.3369	4.0947	0.7101
3.8512	28500	0.3229	4.0495	0.7101
3.9188	29000	0.3442	4.0408	0.7169
3.9864	29500	0.3059	3.9493	0.6959
4.0538	30000	0.3349	4.0431	0.7108
4.1214	30500	0.3266	4.0224	0.7189
4.1889	31000	0.3501	3.9502	0.7169
4.2565	31500	0.3676	3.8903	0.7196
4.3241	32000	0.3191	3.7994	0.7162
4.3916	32500	0.3317	3.7889	0.7182
4.4592	33000	0.3304	3.8661	0.7108
4.5268	33500	0.3332	3.8822	0.7115
4.5943	34000	0.3435	3.7945	0.7088
4.6619	34500	0.317	3.8721	0.7243
4.7295	35000	0.3038	3.8615	0.7209
4.7970	35500	0.3093	3.8360	0.7162
4.8646	36000	0.3309	3.8277	0.7155
4.9322	36500	0.3378	3.7988	0.7128
4.9997	37000	0.311	3.8015	0.7135

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 3.4.1
• Transformers: 4.47.1
• PyTorch: 2.5.1+cu124
• Accelerate: 1.2.1
• 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",
}

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}
}