MedEmbed Biomedical MRL

This is a sentence-transformers model trained. 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
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Language: en
License: apache-2.0

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: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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("potsu-potsu/medembed-base-mrl-train36k")
# Run inference
sentences = [
    'What is Pseudomelanosis duodeni?',
    'Pseudomelanosis duodeni is a rare condition in which dark pigment accumulates in \nmacrophages located in the lamina propria of the duodenal mucosa. Three cases \nare reported here and the literature is reviewed. No clinical association can be \nfound that points clearly to the underlying etiology. Electron probe x-ray \nmicroanalysis was used to study the pigment in macrophage granules in 2 of our \npatients and demonstrated high iron and sulfur content. Iron accumulation in \nferritinlike particles was detected in absorptive cell lysosomes. A possible \nmechanism for the accumulation of absorbed iron by macrophages is considered.',
    'This year marks the 100th anniversary of the deadliest event in human history. \nIn 1918-1919, pandemic influenza appeared nearly simultaneously around the globe \nand caused extraordinary mortality (an estimated 50-100 million deaths) \nassociated with unexpected clinical and epidemiological features. The \ndescendants of the 1918 virus remain today; as endemic influenza viruses, they \ncause significant mortality each year. Although the ability to predict influenza \npandemics remains no better than it was a century ago, numerous scientific \nadvances provide an important head start in limiting severe disease and death \nfrom both current and future influenza viruses: identification and substantial \ncharacterization of the natural history and pathogenesis of the 1918 causative \nvirus itself, as well as hundreds of its viral descendants; development of \nmoderately effective vaccines; improved diagnosis and treatment of \ninfluenza-associated pneumonia; and effective prevention and control measures. \nRemaining challenges include development of vaccines eliciting significantly \nbroader protection (against antigenically different influenza viruses) that can \nprevent or significantly downregulate viral replication; more complete \ncharacterization of natural history and pathogenesis emphasizing the protective \nrole of mucosal immunity; and biomarkers of impending influenza-associated \npneumonia.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

Dataset: dim_768
Evaluated with InformationRetrievalEvaluator with these parameters:
```
{
    "truncate_dim": 768
}
```

Metric	Value
cosine_accuracy@1	0.7341
cosine_accuracy@3	0.8515
cosine_accuracy@5	0.8911
cosine_accuracy@10	0.9279
cosine_precision@1	0.7341
cosine_precision@3	0.6073
cosine_precision@5	0.527
cosine_precision@10	0.4123
cosine_recall@1	0.2179
cosine_recall@3	0.3935
cosine_recall@5	0.489
cosine_recall@10	0.6294
cosine_ndcg@10	0.7018
cosine_mrr@10	0.8017
cosine_map@100	0.6443

Information Retrieval

Dataset: dim_512
Evaluated with InformationRetrievalEvaluator with these parameters:
```
{
    "truncate_dim": 512
}
```

Metric	Value
cosine_accuracy@1	0.7284
cosine_accuracy@3	0.8359
cosine_accuracy@5	0.8868
cosine_accuracy@10	0.918
cosine_precision@1	0.7284
cosine_precision@3	0.5997
cosine_precision@5	0.5267
cosine_precision@10	0.4078
cosine_recall@1	0.2181
cosine_recall@3	0.3851
cosine_recall@5	0.4866
cosine_recall@10	0.6179
cosine_ndcg@10	0.6935
cosine_mrr@10	0.7944
cosine_map@100	0.6375

Information Retrieval

Dataset: dim_256
Evaluated with InformationRetrievalEvaluator with these parameters:
```
{
    "truncate_dim": 256
}
```

Metric	Value
cosine_accuracy@1	0.71
cosine_accuracy@3	0.8331
cosine_accuracy@5	0.8755
cosine_accuracy@10	0.9066
cosine_precision@1	0.71
cosine_precision@3	0.586
cosine_precision@5	0.5092
cosine_precision@10	0.4023
cosine_recall@1	0.2135
cosine_recall@3	0.3792
cosine_recall@5	0.4684
cosine_recall@10	0.6041
cosine_ndcg@10	0.6795
cosine_mrr@10	0.7786
cosine_map@100	0.6206

Information Retrieval

Dataset: dim_128
Evaluated with InformationRetrievalEvaluator with these parameters:
```
{
    "truncate_dim": 128
}
```

Metric	Value
cosine_accuracy@1	0.6931
cosine_accuracy@3	0.8091
cosine_accuracy@5	0.8458
cosine_accuracy@10	0.8911
cosine_precision@1	0.6931
cosine_precision@3	0.5582
cosine_precision@5	0.4885
cosine_precision@10	0.3767
cosine_recall@1	0.2057
cosine_recall@3	0.3584
cosine_recall@5	0.4425
cosine_recall@10	0.5665
cosine_ndcg@10	0.6445
cosine_mrr@10	0.7607
cosine_map@100	0.5817

Information Retrieval

Dataset: dim_64
Evaluated with InformationRetrievalEvaluator with these parameters:
```
{
    "truncate_dim": 64
}
```

Metric	Value
cosine_accuracy@1	0.6294
cosine_accuracy@3	0.7638
cosine_accuracy@5	0.7935
cosine_accuracy@10	0.8458
cosine_precision@1	0.6294
cosine_precision@3	0.5144
cosine_precision@5	0.445
cosine_precision@10	0.3487
cosine_recall@1	0.1813
cosine_recall@3	0.3196
cosine_recall@5	0.3905
cosine_recall@10	0.5098
cosine_ndcg@10	0.5848
cosine_mrr@10	0.7038
cosine_map@100	0.5145

Training Details

Training Dataset

Unnamed Dataset

Size: 36,470 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 6 tokens
mean: 15.86 tokens
max: 32 tokens

min: 31 tokens
mean: 316.54 tokens
max: 512 tokens

	anchor	positive
type	string	string
details	min: 6 tokens mean: 15.86 tokens max: 32 tokens	min: 31 tokens mean: 316.54 tokens max: 512 tokens

Samples:

anchor	positive
`What is the implication of histone lysine methylation in medulloblastoma?`	Recent studies showed frequent mutations in histone H3 lysine 27 (H3K27) demethylases in medulloblastomas of Group 3 and Group 4, suggesting a role for H3K27 methylation in these tumors. Indeed, trimethylated H3K27 (H3K27me3) levels were shown to be higher in Group 3 and 4 tumors compared to WNT and SHH medulloblastomas, also in tumors without detectable mutations in demethylases. Here, we report that polycomb genes, required for H3K27 methylation, are consistently upregulated in Group 3 and 4 tumors. These tumors show high expression of the homeobox transcription factor OTX2. Silencing of OTX2 in D425 medulloblastoma cells resulted in downregulation of polycomb genes such as EZH2, EED, SUZ12 and RBBP4 and upregulation of H3K27 demethylases KDM6A, KDM6B, JARID2 and KDM7A. This was accompanied by decreased H3K27me3 and increased H3K27me1 levels in promoter regions. Strikingly, the decrease of H3K27me3 was most prominent in promoters that bind OTX2. OTX2-bound promoters showe...
`What is the implication of histone lysine methylation in medulloblastoma?`	We used high-resolution SNP genotyping to identify regions of genomic gain and loss in the genomes of 212 medulloblastomas, malignant pediatric brain tumors. We found focal amplifications of 15 known oncogenes and focal deletions of 20 known tumor suppressor genes (TSG), most not previously implicated in medulloblastoma. Notably, we identified previously unknown amplifications and homozygous deletions, including recurrent, mutually exclusive, highly focal genetic events in genes targeting histone lysine methylation, particularly that of histone 3, lysine 9 (H3K9). Post-translational modification of histone proteins is critical for regulation of gene expression, can participate in determination of stem cell fates and has been implicated in carcinogenesis. Consistent with our genetic data, restoration of expression of genes controlling H3K9 methylation greatly diminishes proliferation of medulloblastoma in vitro. Copy number aberrations of genes with critical roles in writing...
`What is the implication of histone lysine methylation in medulloblastoma?`	Recent sequencing efforts have described the mutational landscape of the pediatric brain tumor medulloblastoma. Although MLL2 is among the most frequent somatic single nucleotide variants (SNV), the clinical and biological significance of these mutations remains uncharacterized. Through targeted re-sequencing, we identified mutations of MLL2 in 8 % (14/175) of MBs, the majority of which were loss of function. Notably, we also report mutations affecting the MLL2-binding partner KDM6A, in 4 % (7/175) of tumors. While MLL2 mutations were independent of age, gender, histological subtype, M-stage or molecular subgroup, KDM6A mutations were most commonly identified in Group 4 MBs, and were mutually exclusive with MLL2 mutations. Immunohistochemical staining for H3K4me3 and H3K27me3, the chromatin effectors of MLL2 and KDM6A activity, respectively, demonstrated alterations of the histone code in 24 % (53/220) of MBs across all subgroups. Correlating these MLL2- and KDM6A-driven h...

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 32
per_device_eval_batch_size: 16
gradient_accumulation_steps: 16
learning_rate: 2e-05
num_train_epochs: 4
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: True
tf32: True
load_best_model_at_end: True
optim: adamw_torch_fused
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 16
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 16
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: 4
max_steps: -1
lr_scheduler_type: cosine
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: True
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: True
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}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
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	dim_768_cosine_ndcg@10	dim_512_cosine_ndcg@10	dim_256_cosine_ndcg@10	dim_128_cosine_ndcg@10	dim_64_cosine_ndcg@10
0.1404	10	74.2852	-	-	-	-	-
0.2807	20	53.9914	-	-	-	-	-
0.4211	30	37.5251	-	-	-	-	-
0.5614	40	28.6889	-	-	-	-	-
0.7018	50	23.8582	-	-	-	-	-
0.8421	60	21.6883	-	-	-	-	-
0.9825	70	19.7715	-	-	-	-	-
1.0	72	-	0.7030	0.6976	0.6855	0.6427	0.5803
1.1123	80	16.5108	-	-	-	-	-
1.2526	90	16.5154	-	-	-	-	-
1.3930	100	14.3628	-	-	-	-	-
1.5333	110	15.1679	-	-	-	-	-
1.6737	120	13.5316	-	-	-	-	-
1.8140	130	12.5184	-	-	-	-	-
1.9544	140	13.1961	-	-	-	-	-
2.0	144	-	0.7011	0.6923	0.6803	0.6459	0.5873
2.0842	150	11.1752	-	-	-	-	-
2.2246	160	10.771	-	-	-	-	-
2.3649	170	11.0394	-	-	-	-	-
2.5053	180	10.0241	-	-	-	-	-
2.6456	190	10.862	-	-	-	-	-
2.7860	200	10.39	-	-	-	-	-
2.9263	210	10.6967	-	-	-	-	-
3.0	216	-	0.7014	0.6936	0.6805	0.6450	0.5824
3.0561	220	9.2254	-	-	-	-	-
3.1965	230	9.7925	-	-	-	-	-
3.3368	240	9.6484	-	-	-	-	-
3.4772	250	9.4891	-	-	-	-	-
3.6175	260	9.5589	-	-	-	-	-
3.7579	270	8.773	-	-	-	-	-
3.8982	280	9.3302	-	-	-	-	-
4.0	288	-	0.7018	0.6935	0.6795	0.6445	0.5848

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.12.6
Sentence Transformers: 4.1.0
Transformers: 4.52.4
PyTorch: 2.6.0+cu124
Accelerate: 1.7.0
Datasets: 3.6.0
Tokenizers: 0.21.1

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

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}
}

potsu-potsu
/

medembed-base-mrl-train36k

MedEmbed Biomedical MRL

Model Details

Model Description

Model Sources

Full Model Architecture

Usage

Direct Usage (Sentence Transformers)

Evaluation

Metrics

Information Retrieval

Information Retrieval

Information Retrieval

Information Retrieval

Information Retrieval

Training Details

Training Dataset

Unnamed Dataset

Training Hyperparameters

Non-Default Hyperparameters

All Hyperparameters

Training Logs

Framework Versions

Citation

BibTeX

Sentence Transformers

MatryoshkaLoss

MultipleNegativesRankingLoss

Evaluation results