--- tags: - sentence-transformers - sentence-similarity - feature-extraction - generated_from_trainer - dataset_size:35520 - loss:MultipleNegativesRankingLoss base_model: DeepChem/ChemBERTa-77M-MLM widget: - source_sentence: C[NH+]1CCC(CN2c3ccccc3Sc3ccccc32)C1 sentences: - CC(C)CN(CC(O)C(Cc1ccccc1)NC(=O)OC1COC2OCCC12)S(=O)(=O)c1ccc(N)cc1 - COC(=O)NC(C(=O)NC(Cc1ccccc1)C(O)CN(Cc1ccc(-c2ccccn2)cc1)NC(=O)C(NC(=O)OC)C(C)(C)C)C(C)(C)C - C=C1c2cccc([O-])c2C(=O)C2=C([O-])C3(O)C(=O)C(C(N)=O)=C([O-])C([NH+](C)C)C3C(O)C12 - source_sentence: CC(C)(C)[NH2+]CC(O)COc1ccccc1C1CCCC1 sentences: - C[NH2+]C1C(OC2C(OC3C(O)C(O)C(NC(N)=[NH2+])C(O)C3NC(N)=[NH2+])OC(C)C2(O)C=O)OC(CO)C(O)C1O - CC(C)CNCc1ccc(-c2ccccc2S(=O)(=O)N2CCCC2)cc1 - CC1C[NH+](CC(Cc2ccccc2)C(=O)NCC(=O)[O-])CCC1(C)c1cccc(O)c1 - source_sentence: CC1CC2C3CCC4=CC(=O)C=CC4(C)C3(F)C(O)CC2(C)C1(OC(=O)c1ccccc1)C(=O)CO sentences: - CC1CC=CC=CC=CC=CC(OC2OC(C)C(O)C([NH3+])C2O)CC2OC(O)(CC(O)CC3OC3C=CC(=O)O1)CC(O)C2C(=O)[O-] - C=CC1(C)CC(OC(=O)CSC2CC3CCC(C2)[NH+]3C)C2(C)C(C)CCC3(CCC(=O)C32)C(C)C1O - CC(C)C(CN1CCC(C)(c2cccc(O)c2)C(C)C1)NC(=O)C1Cc2ccc(O)cc2CN1 - source_sentence: CC(C)[NH2+]CC1CCc2cc(CO)c([N+](=O)[O-])cc2N1 sentences: - CC(Cc1cc2c(c(C(N)=O)c1)N(CCCO)CC2)[NH2+]CCOc1ccccc1OCC(F)(F)F - COC(=O)NC(C(=O)NC(Cc1ccccc1)C(O)CN(Cc1ccc(-c2ccccn2)cc1)NC(=O)C(NC(=O)OC)C(C)(C)C)C(C)(C)C - COc1ccccc1Oc1c([N-]S(=O)(=O)c2ccc(C(C)(C)C)cc2)nc(-c2ncccn2)nc1OCCO - source_sentence: COc1ccc(C(=O)CC(=O)c2ccc(C(C)(C)C)cc2)cc1 sentences: - C[N+]1(C)CCC(=C(c2ccccc2)c2ccccc2)CC1 - CC#CCC(C)C(O)C=CC1C(O)CC2CC(=CCCCC(=O)[O-])CC21 - C=C1CC2CCC34CC5OC6C(OC7CCC(CC(=O)CC8C(CC9OC(CCC1O2)CC(C)C9=C)OC(CC(O)CN)C8OC)OC7C6O3)C5O4 pipeline_tag: sentence-similarity library_name: sentence-transformers metrics: - cosine_accuracy model-index: - name: SentenceTransformer based on DeepChem/ChemBERTa-77M-MLM results: - task: type: triplet name: Triplet dataset: name: all dev type: all-dev metrics: - type: cosine_accuracy value: 0.7844594594594595 name: Cosine Accuracy --- # SentenceTransformer based on DeepChem/ChemBERTa-77M-MLM This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [DeepChem/ChemBERTa-77M-MLM](https://huggingface.co/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](https://huggingface.co/DeepChem/ChemBERTa-77M-MLM) - **Maximum Sequence Length:** 512 tokens - **Output Dimensionality:** 384 dimensions - **Similarity Function:** Cosine Similarity ### Model Sources - **Documentation:** [Sentence Transformers Documentation](https://sbert.net) - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers) - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers) ### 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: ```bash pip install -U sentence-transformers ``` Then you can load this model and run inference. ```python from sentence_transformers import SentenceTransformer # Download from the 🤗 Hub model = SentenceTransformer("HassanCS/chemBERTa-tuned-on-ClinTox-4") # Run inference sentences = [ 'COc1ccc(C(=O)CC(=O)c2ccc(C(C)(C)C)cc2)cc1', 'C[N+]1(C)CCC(=C(c2ccccc2)c2ccccc2)CC1', 'C=C1CC2CCC34CC5OC6C(OC7CCC(CC(=O)CC8C(CC9OC(CCC1O2)CC(C)C9=C)OC(CC(O)CN)C8OC)OC7C6O3)C5O4', ] 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](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.TripletEvaluator) | Metric | Value | |:--------------------|:-----------| | **cosine_accuracy** | **0.7845** | ## Training Details ### Training Dataset #### Unnamed Dataset * Size: 35,520 training samples * Columns: anchor, positive, and negative * Approximate statistics based on the first 1000 samples: | | anchor | positive | negative | |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------| | type | string | string | string | | details | | | | * 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: [MultipleNegativesRankingLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters: ```json { "scale": 20.0, "similarity_fct": "cos_sim" } ``` ### 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 | | | | * Samples: | anchor | positive | negative | |:------------------------------------------------------------------------|:-------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------| | CC(C)OC(=O)CCCC=CCC1C(O)CC(O)C1C=CC(O)COc1cccc(C(F)(F)F)c1 | C#CC1(O)CCC2C3CCC4=C(CCC(=O)C4)C3CCC21C | CC(C)CC(NC(=O)C(CCc1ccccc1)NC(=O)CN1CCOCC1)C(=O)NC(Cc1ccccc1)C(=O)NC(CC(C)C)C(=O)C1(C)CO1 | | CC(C)OC(=O)CCCC=CCC1C(O)CC(O)C1C=CC(O)COc1cccc(C(F)(F)F)c1 | C=CC1(C)CC(OC(=O)CSC2CC3CCC(C2)[NH+]3C)C2(C)C(C)CCC3(CCC(=O)C32)C(C)C1O | COC(=O)NC(C(=O)NC(Cc1ccccc1)C(O)CN(Cc1ccc(-c2ccccn2)cc1)NC(=O)C(NC(=O)OC)C(C)(C)C)C(C)(C)C | | CC(C)OC(=O)CCCC=CCC1C(O)CC(O)C1C=CC(O)COc1cccc(C(F)(F)F)c1 | CC(Cc1cc2c(c(C(N)=O)c1)N(CCCO)CC2)[NH2+]CCOc1ccccc1OCC(F)(F)F | CC(C)C1(C(=O)NC2CC(=O)OC2(O)CF)CC(c2nccc3ccccc23)=NO1 | * Loss: [MultipleNegativesRankingLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters: ```json { "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`: 10 - `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`: 10 - `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 | |:------:|:-----:|:-------------:|:---------------:|:-----------------------:| | 0.2252 | 500 | 4.2712 | 3.3651 | 0.45 | | 0.4505 | 1000 | 3.5714 | 2.5580 | 0.6223 | | 0.6757 | 1500 | 3.3655 | 2.5956 | 0.6169 | | 0.9009 | 2000 | 3.2218 | 2.6932 | 0.6493 | | 1.1257 | 2500 | 3.0911 | 2.7852 | 0.6736 | | 1.3509 | 3000 | 3.0007 | 2.7838 | 0.6703 | | 1.5761 | 3500 | 3.0536 | 2.5324 | 0.7311 | | 1.8014 | 4000 | 3.0286 | 2.6623 | 0.6892 | | 2.0261 | 4500 | 2.9539 | 2.6397 | 0.7088 | | 2.2514 | 5000 | 2.9252 | 2.5550 | 0.7419 | | 2.4766 | 5500 | 2.944 | 2.5391 | 0.7419 | | 2.7018 | 6000 | 3.028 | 2.6421 | 0.6919 | | 2.9270 | 6500 | 2.9389 | 2.5931 | 0.7209 | | 3.1518 | 7000 | 2.9006 | 2.6597 | 0.7365 | | 3.3770 | 7500 | 2.9107 | 2.4841 | 0.7709 | | 3.6023 | 8000 | 2.9802 | 2.5128 | 0.7493 | | 3.8275 | 8500 | 2.9498 | 2.5716 | 0.7439 | | 4.0523 | 9000 | 2.9004 | 2.4889 | 0.7669 | | 4.2775 | 9500 | 2.89 | 2.5824 | 0.7453 | | 4.5027 | 10000 | 2.9343 | 2.4388 | 0.7757 | | 4.7279 | 10500 | 2.9666 | 2.4759 | 0.7520 | | 4.9532 | 11000 | 2.9153 | 2.6096 | 0.7399 | | 5.1779 | 11500 | 2.873 | 2.5489 | 0.7520 | | 5.4032 | 12000 | 2.8978 | 2.5579 | 0.7527 | | 5.6284 | 12500 | 2.9576 | 2.5336 | 0.7581 | | 5.8536 | 13000 | 2.93 | 2.4656 | 0.7730 | | 6.0784 | 13500 | 2.8825 | 2.4987 | 0.7730 | | 6.3036 | 14000 | 2.8863 | 2.4866 | 0.7818 | | 6.5288 | 14500 | 2.9221 | 2.4416 | 0.7818 | | 6.7541 | 15000 | 2.9544 | 2.4705 | 0.7622 | | 6.9793 | 15500 | 2.8929 | 2.4991 | 0.7669 | | 7.2041 | 16000 | 2.8656 | 2.5163 | 0.7689 | | 7.4293 | 16500 | 2.8866 | 2.5390 | 0.7689 | | 7.6545 | 17000 | 2.9675 | 2.4476 | 0.7872 | | 7.8797 | 17500 | 2.9094 | 2.4572 | 0.775 | | 8.1045 | 18000 | 2.8743 | 2.4677 | 0.7743 | | 8.3297 | 18500 | 2.8748 | 2.4658 | 0.7872 | | 8.5550 | 19000 | 2.9201 | 2.4412 | 0.7865 | | 8.7802 | 19500 | 2.9437 | 2.4620 | 0.7811 | | 9.0050 | 20000 | 2.881 | 2.4608 | 0.7797 | | 9.2302 | 20500 | 2.8628 | 2.4801 | 0.7770 | | 9.4554 | 21000 | 2.884 | 2.4699 | 0.7831 | | 9.6806 | 21500 | 2.9658 | 2.4519 | 0.7845 | | 9.9059 | 22000 | 2.8991 | 2.4474 | 0.7845 | ### Framework Versions - Python: 3.11.11 - Sentence Transformers: 3.3.1 - 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 ```bibtex @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 ```bibtex @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} } ```