This is a sentence-transformers model finetuned from dmis-lab/biobert-v1.1. 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: dmis-lab/biobert-v1.1
Maximum Sequence Length: 256 tokens
Output Dimensionality: 768 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

HierarchyTransformer(
  (0): Transformer({'max_seq_length': 256, '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'Infectious and parasitic diseases → Mycoses',
    'Infectious and parasitic diseases',
    'Mental illness',
]
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, 0.6610, 0.3361],
#         [0.6610, 1.0000, 0.2730],
#         [0.3361, 0.2730, 1.0000]])

Training Details

Training Dataset

Unnamed Dataset

Size: 148,295 training samples
Columns: child, parent, parent_negative, and child_negative

Approximate statistics based on the first 1000 samples:

	child	parent	parent_negative	child_negative
type	string	string	string	string
details	min: 8 tokens mean: 25.19 tokens max: 65 tokens	min: 4 tokens mean: 16.22 tokens max: 41 tokens	min: 4 tokens mean: 16.94 tokens max: 34 tokens	min: 11 tokens mean: 23.48 tokens max: 65 tokens

Samples:

child	parent	parent_negative	child_negative
`Infectious and parasitic diseases → Bacterial infection`	`Infectious and parasitic diseases`	`Mental illness`	`Diseases of the nervous system and sense organs → Central nervous system infection`
`Infectious and parasitic diseases → Bacterial infection`	`Infectious and parasitic diseases`	`Mental illness`	`Diseases of the digestive system → Intestinal infection`
`Infectious and parasitic diseases → Bacterial infection`	`Infectious and parasitic diseases`	`Mental illness`	`Diseases of the skin and subcutaneous tissue → Skin and subcutaneous tissue infections`

Loss: hierarchy_transformers.losses.symmetric_loss.SymmetricLoss with these parameters:
json
1{ 2 "distance_metric": "PoincareBall(c=0.0013021096820011735).dist and dist0", 3 "HyperbolicChildTriplet": { 4 "weight": 1.0, 5 "distance_metric": "PoincareBall(c=0.0013021096820011735).dist", 6 "margin": 3.0 7 }, 8 "HyperbolicParentTriplet": { 9 "weight": 1.0, 10 "distance_metric": "PoincareBall(c=0.0013021096820011735).dist", 11 "margin": 3.0 12 } 13}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 128
per_device_eval_batch_size: 512
learning_rate: 1e-05
num_train_epochs: 10
warmup_steps: 500
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 128
per_device_eval_batch_size: 512
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: 1e-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: 0.0
warmup_steps: 500
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
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: True
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}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
project: huggingface
trackio_space_id: trackio
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
hub_revision: None
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: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
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
0.0863	100	2.1613
0.1726	200	0.5936
0.2588	300	0.1998
0.3451	400	0.1107
0.4314	500	0.0567
0.5177	600	0.0452
0.6040	700	0.032
0.6903	800	0.0279
0.7765	900	0.0218
0.8628	1000	0.0235
0.9491	1100	0.018
1.0	1159	-
1.0354	1200	0.0192
1.1217	1300	0.0176
1.2079	1400	0.0137
1.2942	1500	0.0119
1.3805	1600	0.0139
1.4668	1700	0.0138
1.5531	1800	0.0123
1.6393	1900	0.0104
1.7256	2000	0.0117
1.8119	2100	0.0097
1.8982	2200	0.0133
1.9845	2300	0.01
2.0	2318	-
2.0708	2400	0.0109
2.1570	2500	0.0074
2.2433	2600	0.0072
2.3296	2700	0.015
2.4159	2800	0.0069
2.5022	2900	0.0107
2.5884	3000	0.0094
2.6747	3100	0.0105
2.7610	3200	0.0095
2.8473	3300	0.0072
2.9336	3400	0.0084
3.0	3477	-
3.0198	3500	0.0104
3.1061	3600	0.0078
3.1924	3700	0.008
3.2787	3800	0.0086
3.3650	3900	0.0085
3.4513	4000	0.0081
3.5375	4100	0.0093
3.6238	4200	0.0107
3.7101	4300	0.008
3.7964	4400	0.0099
3.8827	4500	0.0058
3.9689	4600	0.0084
4.0	4636	-
4.0552	4700	0.01
4.1415	4800	0.0053
4.2278	4900	0.0075
4.3141	5000	0.0077
4.4003	5100	0.0065
4.4866	5200	0.0089
4.5729	5300	0.0082
4.6592	5400	0.0093
4.7455	5500	0.0076
4.8318	5600	0.0095
4.9180	5700	0.0078
5.0	5795	-
5.0043	5800	0.0055
5.0906	5900	0.0061
5.1769	6000	0.005
5.2632	6100	0.0075
5.3494	6200	0.0079
5.4357	6300	0.006
5.5220	6400	0.0095
5.6083	6500	0.0099
5.6946	6600	0.0084
5.7808	6700	0.008
5.8671	6800	0.0064
5.9534	6900	0.0097
6.0	6954	-
6.0397	7000	0.0063
6.1260	7100	0.0069
6.2123	7200	0.0095
6.2985	7300	0.0067
6.3848	7400	0.0056
6.4711	7500	0.0074
6.5574	7600	0.0086
6.6437	7700	0.0072
6.7299	7800	0.0065
6.8162	7900	0.0052
6.9025	8000	0.0101
6.9888	8100	0.0086
7.0	8113	-
7.0751	8200	0.0065
7.1613	8300	0.0106
7.2476	8400	0.0049
7.3339	8500	0.0074
7.4202	8600	0.0065
7.5065	8700	0.004
7.5928	8800	0.0075
7.6790	8900	0.009
7.7653	9000	0.0059
7.8516	9100	0.0063
7.9379	9200	0.0095
8.0	9272	-
8.0242	9300	0.0082
8.1104	9400	0.0067
8.1967	9500	0.0063
8.2830	9600	0.0071
8.3693	9700	0.0064
8.4556	9800	0.0072
8.5418	9900	0.0059
8.6281	10000	0.0085
8.7144	10100	0.0083
8.8007	10200	0.0046
8.8870	10300	0.0055
8.9733	10400	0.008
9.0	10431	-
9.0595	10500	0.0066
9.1458	10600	0.0068
9.2321	10700	0.0093
9.3184	10800	0.0067
9.4047	10900	0.0054
9.4909	11000	0.0079
9.5772	11100	0.0052
9.6635	11200	0.0073
9.7498	11300	0.0088
9.8361	11400	0.005
9.9223	11500	0.0069
10.0	11590	-

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.13
Sentence Transformers: 5.1.2
Transformers: 4.57.1
PyTorch: 2.9.0+cu128
Accelerate: 1.11.0
Datasets: 4.3.0
Tokenizers: 0.22.1

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",
}

SymmetricLoss