This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5. 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: BAAI/bge-base-en-v1.5
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 tokens
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

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': True}) 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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("Nutanix/bge-base-mbpp")
# Run inference
sentences = [
    'Write a function to find sum and average of first n natural numbers.',
    'def sum_average(number):\r\n total = 0\r\n for value in range(1, number + 1):\r\n    total = total + value\r\n average = total / number\r\n return (total,average)',
    'def long_words(n, str):\r\n    word_len = []\r\n    txt = str.split(" ")\r\n    for x in txt:\r\n        if len(x) > n:\r\n            word_len.append(x)\r\n    return word_len\t',
]
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

Triplet

Dataset: sts-dev
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9971
dot_accuracy	0.0028
manhattan_accuracy	0.9961
euclidean_accuracy	0.9971
max_accuracy	0.9971

Training Details

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 1
bf16: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
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
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: 1
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0
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
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: 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, '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_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	sts-dev_max_accuracy
0.0050	100	4.3364	-
0.0101	200	4.122	-
0.0151	300	4.0825	-
0.0202	400	4.0381	-
0.0252	500	4.015	-
0.0302	600	3.9996	-
0.0353	700	3.9567	-
0.0403	800	3.9593	-
0.0453	900	3.9456	-
0.0504	1000	3.938	-
0.0554	1100	3.933	-
0.0605	1200	3.905	-
0.0655	1300	3.906	-
0.0705	1400	3.9073	-
0.0756	1500	3.9193	-
0.0806	1600	3.9016	-
0.0857	1700	3.8899	-
0.0907	1800	3.9	-
0.0957	1900	3.8983	-
0.1008	2000	3.876	-
0.1058	2100	3.9001	-
0.1109	2200	3.8818	-
0.1159	2300	3.8788	-
0.1209	2400	3.8815	-
0.1260	2500	3.8664	-
0.1310	2600	3.854	-
0.1360	2700	3.8674	-
0.1411	2800	3.8525	-
0.1461	2900	3.8733	-
0.1512	3000	3.8538	-
0.1562	3100	3.8348	-
0.1612	3200	3.8378	-
0.1663	3300	3.8504	-
0.1713	3400	3.8409	-
0.1764	3500	3.8436	-
0.1814	3600	3.8422	-
0.1864	3700	3.8629	-
0.1915	3800	3.8589	-
0.1965	3900	3.8572	-
0.2016	4000	3.8309	-
0.2066	4100	3.8465	-
0.2116	4200	3.8311	-
0.2167	4300	3.8124	-
0.2217	4400	3.8412	-
0.2267	4500	3.8228	-
0.2318	4600	3.8012	-
0.2368	4700	3.8185	-
0.2419	4800	3.8242	-
0.2469	4900	3.7917	-
0.2519	5000	3.8022	-
0.2570	5100	3.7991	-
0.2620	5200	3.7943	-
0.2671	5300	3.7874	-
0.2721	5400	3.7987	-
0.2771	5500	3.7982	-
0.2822	5600	3.7789	-
0.2872	5700	3.7837	-
0.2923	5800	3.7762	-
0.2973	5900	3.7854	-
0.3023	6000	3.7719	-
0.3074	6100	3.7925	-
0.3124	6200	3.7795	-
0.3174	6300	3.7725	-
0.3225	6400	3.7897	-
0.3275	6500	3.773	-
0.3326	6600	3.7803	-
0.3376	6700	3.7476	-
0.3426	6800	3.7585	-
0.3477	6900	3.7426	-
0.3527	7000	3.7529	-
0.3578	7100	3.7745	-
0.3628	7200	3.7771	-
0.3678	7300	3.7598	-
0.3729	7400	3.7428	-
0.3779	7500	3.7409	-
0.3829	7600	3.7569	-
0.3880	7700	3.7517	-
0.3930	7800	3.7484	-
0.3981	7900	3.7415	-
0.4031	8000	3.7228	-
0.4081	8100	3.7569	-
0.4132	8200	3.7421	-
0.4182	8300	3.7233	-
0.4233	8400	3.72	-
0.4283	8500	3.7431	-
0.4333	8600	3.7258	-
0.4384	8700	3.73	-
0.4434	8800	3.7286	-
0.4485	8900	3.7487	-
0.4535	9000	3.7359	-
0.4585	9100	3.7387	-
0.4636	9200	3.7135	-
0.4686	9300	3.7219	-
0.4736	9400	3.7189	-
0.4787	9500	3.7234	-
0.4837	9600	3.7333	-
0.4888	9700	3.7027	-
0.4938	9800	3.7358	-
0.4988	9900	3.6959	-
0.5039	10000	3.7051	-
0.5089	10100	3.7205	-
0.5140	10200	3.711	-
0.5190	10300	3.6898	-
0.5240	10400	3.7103	-
0.5291	10500	3.695	-
0.5341	10600	3.7108	-
0.5392	10700	3.7226	-
0.5442	10800	3.7004	-
0.5492	10900	3.736	-
0.5543	11000	3.7135	-
0.5593	11100	3.7148	-
0.5643	11200	3.7285	-
0.5694	11300	3.694	-
0.5744	11400	3.6913	-
0.5795	11500	3.69	-
0.5845	11600	3.7249	-
0.5895	11700	3.6907	-
0.5946	11800	3.7135	-
0.5996	11900	3.7172	-
0.6047	12000	3.7087	-
0.6097	12100	3.7045	-
0.6147	12200	3.7043	-
0.6198	12300	3.693	-
0.6248	12400	3.6982	-
0.6298	12500	3.6922	-
0.6349	12600	3.6857	-
0.6399	12700	3.6834	-
0.6450	12800	3.7052	-
0.6500	12900	3.6935	-
0.6550	13000	3.6736	-
0.6601	13100	3.7026	-
0.6651	13200	3.6846	-
0.6702	13300	3.704	-
0.6752	13400	3.6818	-
0.6802	13500	3.7075	-
0.6853	13600	3.6688	-
0.6903	13700	3.6933	-
0.6954	13800	3.6971	-
0.7004	13900	3.6785	-
0.7054	14000	3.7088	-
0.7105	14100	3.7127	-
0.7155	14200	3.6996	-
0.7205	14300	3.6901	-
0.7256	14400	3.6914	-
0.7306	14500	3.6659	-
0.7357	14600	3.6859	-
0.7407	14700	3.68	-
0.7457	14800	3.6874	-
0.7508	14900	3.6854	-
0.7558	15000	3.671	-
0.7609	15100	3.6909	-
0.7659	15200	3.7014	-
0.7709	15300	3.6828	-
0.7760	15400	3.6773	-
0.7810	15500	3.6863	-
0.7861	15600	3.6892	-
0.7911	15700	3.6864	-
0.7961	15800	3.6586	-
0.8012	15900	3.6639	-
0.8062	16000	3.6843	-
0.8112	16100	3.6865	-
0.8163	16200	3.678	-
0.8213	16300	3.6825	-
0.8264	16400	3.7068	-
0.8314	16500	3.6886	-
0.8364	16600	3.6905	-
0.8415	16700	3.6905	-
0.8465	16800	3.6677	-
0.8516	16900	3.684	-
0.8566	17000	3.6872	-
0.8616	17100	3.6849	-
0.8667	17200	3.662	-
0.8717	17300	3.6887	-
0.8768	17400	3.6999	-
0.8818	17500	3.6916	-
0.8868	17600	3.6853	-
0.8919	17700	3.6971	-
0.8969	17800	3.6846	-
0.9019	17900	3.6701	-
0.9070	18000	3.6911	-
0.9120	18100	3.7021	-
0.9171	18200	3.6851	-
0.9221	18300	3.6924	-
0.9271	18400	3.6644	-
0.9322	18500	3.6674	-
0.9372	18600	3.6962	-
0.9423	18700	3.6759	-
0.9473	18800	3.6839	-
0.9523	18900	3.6822	-
0.9574	19000	3.6947	-
0.9624	19100	3.6589	-
0.9674	19200	3.6817	-
0.9725	19300	3.6754	-
0.9775	19400	3.6947	-
0.9826	19500	3.6785	-
0.9876	19600	3.6776	-
0.9926	19700	3.6791	-
0.9977	19800	3.6795	-
1.0	19846	-	0.9971

Framework Versions

Python: 3.10.14
Sentence Transformers: 3.0.1
Transformers: 4.40.0
PyTorch: 2.3.0+cu121
Accelerate: 0.33.0
Datasets: 2.20.0
Tokenizers: 0.19.1

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",
}

TripletLoss

bibtex

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification}, 
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}