SentenceTransformer based on sentence-transformers/paraphrase-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-MiniLM-L6-v2. 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: sentence-transformers/paraphrase-MiniLM-L6-v2
Maximum Sequence Length: 128 tokens
Output Dimensionality: 384 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

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("hskwon7/paraphrase-MiniLM-L6-v2-ft-for-etf-semantic-search")
# Run inference
sentences = [
    'Low Return Moderate Risk ETF',
    'Arrow Dow Jones Global Yield ET. Arrow Dow Jones Global Yield ETF. Issuer: ArrowShares. Category: Global Allocation. Type: Exchange Traded Fund. Position: buy, bull, long. Return: Low Return. Risk: Moderate Risk. Expense Ratio: High Expense Ratio. Dividend Yield: High Dividend Yield. The fund uses a "passive" or "indexing" investment approach to seek to track the price and yield performance of the index. It invests at least 80% of its total assets in the component securities of the index (or depositary receipts representing those securities). The index seeks to identify the 150 highest yielding investable securities in the world within three "asset classes.". Holdings: Komercni Banka AS (KOMB.PR): 1.4%, Fortum Oyj (FORTUM.HE): 1.3%, Altria Group Inc (MO): 1.3%, Enagas SA (ENG.MC): 1.2%, Phoenix Group Holdings PLC (PHNX.L): 1.2%, British American Tobacco PLC (BATS.L): 1.2%, Medical Properties Trust Inc (MPW): 1.2%, Kumba Iron Ore Ltd (KIO.JO): 1.2%, PT Bukit Asam Tbk Registered Shs Series -B- (PTBA.JK): 1.1%, Exxaro Resources Ltd (EXX.JO): 1.1%',
    'Innovator U.S. Equity Power Buf. Innovator U.S. Equity Power Buffer ETF - August. Issuer: Innovator ETFs. Category: Defined Outcome. Type: Exchange Traded Fund. Position: buy, bull, long. Return: Moderate Return. Risk: Low Risk. Expense Ratio: High Expense Ratio. Dividend Yield: Low Dividend Yield. The fund invests under normal circumstances, at least 80% of its net assets (plus any borrowings for investment purposes) in investments that provide exposure to the SPDR ® S&P 500 ® ETF Trust (the “Underlying ETF”). FLEX Options are exchange-traded option contracts with uniquely customizable terms. It is non-diversified.',
]
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]

Training Details

Training Dataset

Unnamed Dataset

Size: 91,099 training samples
Columns: sentence_0, sentence_1, and label
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1 label
type string string float
details
min: 5 tokens
mean: 10.9 tokens
max: 21 tokens

min: 104 tokens
mean: 127.74 tokens
max: 128 tokens

min: 1.0
mean: 1.0
max: 1.0

	sentence_0	sentence_1	label
type	string	string	float
details	min: 5 tokens mean: 10.9 tokens max: 21 tokens	min: 104 tokens mean: 127.74 tokens max: 128 tokens	min: 1.0 mean: 1.0 max: 1.0

Samples:

sentence_0	sentence_1	label
`Defined Outcome Low Risk ETF`	Innovator U.S. Equity Buffer ET. Innovator U.S. Equity Buffer ETF - June. Issuer: Innovator ETFs. Category: Defined Outcome. Type: Exchange Traded Fund. Position: buy, bull, long. Return: Moderate Return. Risk: Low Risk. Expense Ratio: High Expense Ratio. Dividend Yield: Low Dividend Yield. The fund has adopted a policy pursuant to Rule 35d-1 under the 1940 Act to invest, under normal circumstances, at least 80% of its net assets (plus any borrowings for investment purposes) in investments that provide exposure to the SPDR® S&P 500® ETF Trust (the “Underlying ETF”). It is non-diversified.	`1.0`
`Miscellaneous Region Moderate Return Moderate Risk Low Expense Ratio Moderate Dividend Yield ETF`	Franklin FTSE Germany ETF. Franklin FTSE Germany ETF. Issuer: Franklin Templeton Investments. Category: Miscellaneous Region. Type: Exchange Traded Fund. Position: buy, bull, long. Return: Moderate Return. Risk: Moderate Risk. Expense Ratio: Low Expense Ratio. Dividend Yield: Moderate Dividend Yield. Under normal market conditions, the fund invests at least 80% of its assets in the component securities of the FTSE Germany Capped Index and in depositary receipts representing such securities. The FTSE Germany Capped Index is based on the FTSE Germany Index and is designed to measure the performance of German large- and mid-capitalization stocks. The fund is non-diversified.. Holdings: SAP SE (SAP.DE): 17.8%, Siemens AG (SIE.DE): 10.0%, Allianz SE (ALV.DE): 7.9%, Deutsche Telekom AG (DTE.DE): 7.0%, Mercedes-Benz Group AG (MBG.DE): 3.1%, Infineon Technologies AG (IFX.DE): 2.8%, Deutsche Boerse AG (DB1.DE): 2.8%, Rheinmetall AG (RHM.DE): 2.6%, Basf SE (BAS.DE): 2.6%	`1.0`
`Moderate Risk ETF`	Rareview Dynamic Fixed Income E. Rareview Dynamic Fixed Income ETF. Issuer: Rareview Capital. Category: Bond. Type: Exchange Traded Fund. Position: buy, bull, long. Return: Moderate Return. Risk: Moderate Risk. Expense Ratio: High Expense Ratio. Dividend Yield: High Dividend Yield. Under normal market conditions, the fund will invest at least 80% of its net assets (plus any borrowings for investment purposes) in fixed income closed-end funds trading at a discount or premium to their underlying net asset value and that pay regular periodic cash distributions. Through its investments in closed-end funds that hold non-U.S. fixed income securities, the fund may invest indirectly in foreign securities, including securities of issuers located in emerging markets.. Holdings: SPDR® Blmbg 1-3 Mth T-Bill ETF (BIL): 16.8%, Western Asset Emerg Mkts Debt (EMD): 5.4%, MS Emerging Markets Domestic (EDD): 5.0%, Allspring Inc Opp (EAD): 3.8%, Templeton Emerging Markets Income (TEI): 3.5%, abrdn Asia-Pa...	`1.0`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 2
disable_tqdm: True
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
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
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
num_train_epochs: 2
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
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
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: True
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}
tp_size: 0
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
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: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss
0.0878	500	1.3374
0.1756	1000	0.8677
0.2634	1500	0.785
0.3512	2000	0.7487
0.4391	2500	0.7374
0.5269	3000	0.713
0.6147	3500	0.7142
0.7025	4000	0.6895
0.7903	4500	0.6907
0.8781	5000	0.6976
0.9659	5500	0.6975
1.0537	6000	0.6712
1.1416	6500	0.6661
1.2294	7000	0.6708
1.3172	7500	0.6655
1.4050	8000	0.6627
1.4928	8500	0.6601
1.5806	9000	0.6665
1.6684	9500	0.6649
1.7562	10000	0.6675
1.8440	10500	0.6639
1.9319	11000	0.6634

Framework Versions

Python: 3.11.12
Sentence Transformers: 4.1.0
Transformers: 4.51.3
PyTorch: 2.6.0+cu124
Accelerate: 1.6.0
Datasets: 2.14.4
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",
}

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}
}

hskwon7
/

paraphrase-MiniLM-L6-v2-ft-for-etf-semantic-search