This is a sentence-transformers model finetuned from BAAI/bge-m3 on the json dataset. It maps sentences & paragraphs to a 1024-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: BAAI/bge-m3
Maximum Sequence Length: 8192 tokens
Output Dimensionality: 1024 tokens
Similarity Function: Cosine Similarity
Training Dataset:
- json

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': 8192, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("FareedKhan/BAAI_bge-m3_FareedKhan_prime_synthetic_data_2k_2_4")
# Run inference
sentences = [
    "\nRNASE6, also known by aliases such as RAD1, RNS6, and RNasek6, functions as a member of the ribonuclease A superfamily. Specifically identified via the NCBI gene/protein database, this protein is related to the antimicrobial peptides pathway, showcasing broad-spectrum antimicrobial activity against pathogenic bacteria in the urinary tract. The provided gene summary emphasizes its role in the urinary tract, highlighting its enzymatic function and broad antimicrobial capability.\n\nWith a genomic position spanning from 20781268 to 20782467 on chromosome 14, the RNASE6 gene encodes a protein named ribonuclease A family member k6. The protein's interactions with cellular and molecular functions are integral to its role, including its interaction with molecular functions like ribonuclease activity and endonuclease activity, as well as its involvement in nucleic acid binding.\n\nRNASE6's involvement in biological",
    'Identify genes or proteins linked to encephalopathy that are involved in the Antimicrobial peptides pathway and have interactions with molecular functions associated with ribonuclease activity.',
    'Identify the common gene targets that interact with both N-(3,5-dibromo-4-hydroxyphenyl)benzamide and Diflunisal.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# 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

Metric	Value
cosine_accuracy@1	0.4653
cosine_accuracy@3	0.5149
cosine_accuracy@5	0.5297
cosine_accuracy@10	0.5644
cosine_precision@1	0.4653
cosine_precision@3	0.1716
cosine_precision@5	0.1059
cosine_precision@10	0.0564
cosine_recall@1	0.4653
cosine_recall@3	0.5149
cosine_recall@5	0.5297
cosine_recall@10	0.5644
cosine_ndcg@10	0.5137
cosine_mrr@10	0.4977
cosine_map@100	0.5031

Training Details

Training Dataset

json

Dataset: json
Size: 1,814 training samples
Columns: positive and anchor
Approximate statistics based on the first 1000 samples:
positive anchor
type string string
details
min: 6 tokens
mean: 269.24 tokens
max: 810 tokens
min: 16 tokens
mean: 40.47 tokens
max: 127 tokens

Samples:

positive	anchor
[object Object][object Object]The list you've provided seems to be a collection of various chemical compounds and drug names. These chemicals and drugs are used across multiple fields such as pharmacology, microbiology, and biochemistry. Here are some explanations and examples for the categories:[object Object][object Object]1. [object Object]:[object Object] - [object Object]: An antibiotic used to treat serious bacterial infections.[object Object] - [object Object]: Used for the treatment of blood clots, particularly in people with atrial fibrillation.[object Object] - [object Object]: An anti-inflammatory medication.[object Object] - [object Object]: These are also antibiotics that are effective against various types of microorganisms.[object Object][object Object]2. [object Object]:[object Object] - [object Object]: These are usually found in plants and can have estrogen-like actions, sometimes used in dietary supplements.[object Object] - [object Object]: Antibiotics used for specific infections in humans.[object Object]	`Is there a list of medications that not only exhibit a synergistic effect when combined with Reteplase for managing a common condition but also share Warfarin's mechanism of inhibiting vitamin K reductase, leading to a depletion of reduced vitamin K levels?`
[object Object]Pterygium, a benign conjunctival neoplasm subtype, is associated with the BICD2 gene/protein, according to the document. This wedge-shaped fibrovascular lesion, originating from the bulbar conjunctiva and extending to the cornea, is linked to pterygium due to chronic exposure to solar ultraviolet radiation, heat, and dust. Pterygium may cause severe vision loss and studies suggest it could be related to neoplastic proliferation and be a stem cell disorder. The document also connects pterygium to the IFNA2 gene/protein alongside BICD2, and it is classified among familial pterygium of the conjunctiva, conjunctival pterygium, benign neoplasm of cornea, and benign conjunctival neoplasm.	`Which benign conjunctival neoplasm subtype is associated with the BICD2 gene/protein?`
[object Object]The query you provided is a complex request that seems to be aimed at finding specific genes in the human genome. Here's a breakdown of the components of the query:[object Object][object Object]1. [object Object]: It's searching a database of human long non-coding RNAs (lncRNAs) using an SQL-like query, which is a powerful way to filter through large datasets to find specific genes or transcripts.[object Object][object Object]2. [object Object]: A few columns are explicitly mentioned:[object Object] - [object Object]: This likely refers to the gene identifier.[object Object] - [object Object]: Identifies which chromosome the gene is located on.[object Object] - [object Object]: The starting position of the gene on the chromosome.[object Object] - [object Object]: The ending position of the gene on the chromosome.[object Object][object Object]3. [object Object]: Some columns are related to the DNA sequence of the gene:[object Object] - [object Object]: The full DNA sequence of the gene.[object Object] - [object Object]: Length of the DNA sequence.[object Object][object Object]4. [object Object]: It looks for annotations related to lncRNAs:[object Object] - [object Object]: The type of lncRNA.[object Object] - [object Object]: A specific name for the lncRNA.[object Object][object Object]5. [object Object]:[object Object] - [object Object]: Checks for overlapping or interacting lncRNAs with the query lncRNA.[object Object][object Object]6. [object Object]:[object Object] - [object Object]: Name of the microRNA.[object Object] - [object Object]: Type of the microRNA.[object Object][object Object]7. [object Object]: It also seems to include a list of other gene symbols in the form [object Object]. These might be part of a query to filter or compare against a list of lncRNAs or genes.[object Object][object Object]### How It Works:[object Object]This query is highly specific and might be used by bioinformaticians or researchers to explore interactions between different types of RNA molecules, particularly long non-coding RNAs and microRNAs, within the context of human genetics. It allows for detailed inquiries into gene positions, interactions, and annotations, which are crucial for understanding regulatory processes	`Which anatomical structures lack the expression of genes or proteins that play a role in the organization of microvilli?`

Loss: MatryoshkaLoss with these parameters:
json
1{ 2 "loss": "MultipleNegativesRankingLoss", 3 "matryoshka_dims": [ 4 768 5 ], 6 "matryoshka_weights": [ 7 1 8 ], 9 "n_dims_per_step": -1 10}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 4
learning_rate: 1e-05
num_train_epochs: 2
warmup_ratio: 0.1
bf16: True
tf32: False
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: 4
per_device_eval_batch_size: 8
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: 2
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: True
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: False
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
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: False
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
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
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_768_cosine_map@100
0	0	-	0.4068
0.0220	10	0.6723	-
0.0441	20	0.4606	-
0.0661	30	0.4175	-
0.0881	40	0.5941	-
0.1101	50	0.6731	-
0.1322	60	0.6091	-
0.1542	70	0.2194	-
0.1762	80	0.4262	-
0.1982	90	0.5397	-
0.2203	100	0.3848	-
0.2423	110	0.4412	-
0.2643	120	0.4608	-
0.2863	130	0.3824	-
0.3084	140	0.3241	-
0.3304	150	0.3768	-
0.3524	160	0.4086	-
0.3744	170	0.1928	-
0.3965	180	0.5528	-
0.4185	190	0.4628	-
0.4405	200	0.1467	-
0.4626	210	0.3845	-
0.4846	220	0.275	-
0.5066	230	0.3052	-
0.5286	240	0.3463	-
0.5507	250	0.3588	-
0.5727	260	0.2317	-
0.5947	270	0.14	-
0.6167	280	0.5766	-
0.6388	290	0.4077	-
0.6608	300	0.4412	-
0.6828	310	0.3011	-
0.7048	320	0.1437	-
0.7269	330	0.3659	-
0.7489	340	0.5164	-
0.7709	350	0.4847	-
0.7930	360	0.1533	-
0.8150	370	0.4391	-
0.8370	380	0.195	-
0.8590	390	0.1703	-
0.8811	400	0.5926	-
0.9031	410	0.193	-
0.9251	420	0.1602	-
0.9471	430	0.3054	-
0.9692	440	0.1246	-
0.9912	450	0.1128	-
1.0	454	-	0.4923
1.0132	460	0.3826	-
1.0352	470	0.3034	-
1.0573	480	0.416	-
1.0793	490	0.1682	-
1.1013	500	0.2183	-
1.1233	510	0.2287	-
1.1454	520	0.0765	-
1.1674	530	0.2437	-
1.1894	540	0.0393	-
1.2115	550	0.2171	-
1.2335	560	0.0611	-
1.2555	570	0.0328	-
1.2775	580	0.168	-
1.2996	590	0.1535	-
1.3216	600	0.1937	-
1.3436	610	0.3172	-
1.3656	620	0.3847	-
1.3877	630	0.0818	-
1.4097	640	0.3286	-
1.4317	650	0.3761	-
1.4537	660	0.1369	-
1.4758	670	0.1861	-
1.4978	680	0.2745	-
1.5198	690	0.3078	-
1.5419	700	0.205	-
1.5639	710	0.2251	-
1.5859	720	0.2627	-
1.6079	730	0.2258	-
1.6300	740	0.0863	-
1.6520	750	0.2577	-
1.6740	760	0.1943	-
1.6960	770	0.1056	-
1.7181	780	0.1382	-
1.7401	790	0.0169	-
1.7621	800	0.2602	-
1.7841	810	0.1094	-
1.8062	820	0.1261	-
1.8282	830	0.2647	-
1.8502	840	0.0952	-
1.8722	850	0.13	-
1.8943	860	0.0819	-
1.9163	870	0.1194	-
1.9383	880	0.2118	-
1.9604	890	0.0667	-
1.9824	900	0.2501	-
2.0	908	-	0.5031

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.10
Sentence Transformers: 3.1.1
Transformers: 4.45.1
PyTorch: 2.2.1+cu121
Accelerate: 0.34.2
Datasets: 3.0.1
Tokenizers: 0.20.0

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",
}

MatryoshkaLoss

bibtex

@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

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}
}

	positive	anchor
type	string	string
details	min: 6 tokens mean: 269.24 tokens max: 810 tokens	min: 16 tokens mean: 40.47 tokens max: 127 tokens