SentenceTransformer based on BAAI/bge-small-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-small-en-v1.5 on the csv dataset. 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: BAAI/bge-small-en-v1.5
Maximum Sequence Length: 512 tokens
Output Dimensionality: 384 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- csv

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': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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("sentence_transformers_model_id")
# Run inference
sentences = [
    'Represent this sentence for searching relevant passages: What are some effects of maternal iron deficiency on adult male offspring development?',
    'We found three differentially abundant taxonomic classes in the IDD group using an LDA effect size calculation with an LDA score higher than 4.0. The results showed that the Bacteroidaceae genus Bacteroides and Lachnospiraceae genus Marvinbryantia were significantly increased in rats in the IDD group compared to rats in the other groups (C). \nIn this study, we showed that maternal iron deficiency may program and alter adult male offspring development with regard to spatial learning and memory, dorsal hippocampus BDNF expression, gut microbiota, and SCFA concentrations. Our results showed that the adult male offspring of rats that were fed a low-iron diet before pregnancy and throughout the lactation period had (1) spatial deficits via a Morris water maze evaluation; (2) decreased dorsal hippocampal BDNF mRNA and protein concentrations accompanied by a low TrkB abundance; (3) a decreased plasma acetate concentration without changes in butyrate and propionate concentrations; (4) enrichment of the Bacteroidaceae genus Bacteroides and Lachnospiraceae genus Marvinbryantia.',
    'Parents report encouraging their children to engage in “healthy” lifestyle choices, including making alterations to diet, physical activity (PA), and sleep behavior, which may (1) help parents feel more in control over the impact of the condition, and (2) allow them gain a more positive outlook on the future. Unfortunately, even in the adult MS literature, there is insufficient evidence to make clinical recommendations regarding lifestyle modifications. Improving the body of literature on modifiable lifestyle factors in pediatric MS with the goal of creating guidelines that will help POMS patients and their parents deal with these difficult decisions is needed. \nOur objective in this manuscript is to summarize and identify gaps in current research on modifiable lifestyle factors and pediatric MS. Two questions guided this review: (1) Which modifiable lifestyle factors have been investigated in the context of POMS? And (2) which factors have been shown to play a role in the risk of POMS, disease course, or quality of life? \nWe used the Arksey and O’Malley framework to guide this review.',
]
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

Information Retrieval

Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5854
cosine_accuracy@3	0.7196
cosine_accuracy@5	0.7635
cosine_accuracy@10	0.8084
cosine_precision@1	0.5854
cosine_precision@3	0.2399
cosine_precision@5	0.1527
cosine_precision@10	0.0808
cosine_recall@1	0.5854
cosine_recall@3	0.7196
cosine_recall@5	0.7635
cosine_recall@10	0.8084
cosine_ndcg@10	0.6971
cosine_mrr@10	0.6615
cosine_map@100	0.6663

Training Details

Training Dataset

csv

Dataset: csv
Size: 650,596 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 16 tokens
mean: 26.5 tokens
max: 65 tokens

min: 25 tokens
mean: 229.67 tokens
max: 512 tokens

	anchor	positive
type	string	string
details	min: 16 tokens mean: 26.5 tokens max: 65 tokens	min: 25 tokens mean: 229.67 tokens max: 512 tokens

Samples:

anchor	positive
`Represent this sentence for searching relevant passages: What conditions are excluded as secondary causes of hypercholesterolemia?`	In addition, no abnormalities were revealed under physical examination. The exclusion criteria comprised secondary causes of hypercholesterolemia, including hypothyroidism, kidney diseases, poorly-controlled diabetes, cholestasis or the use of drugs impairing lipid metabolism. The investigation was approved by the Bioethics Committee of the Medical University of Lodz (RNN/191/21/KE). Informed consent was obtained from all participants. All methods were carried out in accordance with relevant guidelines and regulations. All participants were interviewed for their personal history of diabetes, hypertension, smoking, cardiovascular disease, pharmacological treatment, family history of hypercholesterolemia and cardiovascular disease. During the same visit, a physical examination for the presence of corneal arcus and tendon xanthomas was performed. In both the control and research groups, peripheral blood mononuclear cells (PBMCs) and serum were isolated from peripheral whole blood. All...
`Represent this sentence for searching relevant passages: What type of mannose linkage in side chains has the highest impact on antibody response?`	On the other hand, side chains with β-(1→2)-linked mannose residues, which have the highest impact on antibody response , were found only in Candida spp.. The oligomannoside sequence within S. cerevisiae mannan corresponding to antibodies associated with Crohn’s disease was assigned to be the following mannotetraoside: Man(1→3)Man(1→2)Man(1→2)Man , which is illustrated in. Therefore, the corresponding oligosaccharide 1 was selected in this study as a basis for the creation of structurally related glycoarray. Ligands 2 and 3 stem from 1 after formally replacing the terminal α-(1→3)-mannoside fragment with α-(1→2)- and β-(1→2)-mannoside units, respectively. Additional glycosylation of ligand 1 leads to the formation of ligands 4 and 5.
`Represent this sentence for searching relevant passages: How do fluctuations in nest temperature affect bumblebee colonies in aboveground nest boxes?`	Impairments to colony function, as a result a sublethal environmental stressors, are linked with reduced colony success , therefore, combined increases in worker abandonment and reduced offspring production may act to have the greatest impact on bumblebee colony success under chronic heat stress. The results obtained from our laboratory study inform about the capacity of bumblebee colonies to cope with chronic warm temperatures, but there are several distinctions when transposed to natural settings. Conditions used correspond more to surface or aboveground nesting that provide minor buffering from the environment. Underground nest sites are the most frequently observed nesting strategies across multiple bumblebee species, including B. impatiens. However, surface or aboveground nest sites combined are almost as frequently reported for natural settings and even more frequent when nesting in artificial nest such as human made structures. Aboveground temperatures can cause wide fluctuatio...

Loss: CachedGISTEmbedLoss with these parameters:

{'guide': SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: NewModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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})
  (2): Dense({'in_features': 1024, 'out_features': 1024, 'bias': True, 'activation_function': 'torch.nn.modules.linear.Identity'})
), 'temperature': 0.01}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 32768
num_train_epochs: 8
lr_scheduler_type: cosine
warmup_ratio: 0.1
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: 32768
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: 5e-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: 8
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: 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: 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