SentenceTransformer based on google/embeddinggemma-2

This is a sentence-transformers model finetuned from google/embeddinggemma-2. It maps inputs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: google/embeddinggemma-2
Maximum Sequence Length: 4096 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Supported Modalities: Text, Image, Audio, Video, Message

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}, 'image': {'method': 'forward', 'method_output_name': 'last_hidden_state'}, 'audio': {'method': 'forward', 'method_output_name': 'last_hidden_state'}, 'video': {'method': 'forward', 'method_output_name': 'last_hidden_state'}, 'message': {'method': 'forward', 'method_output_name': 'last_hidden_state', 'format': 'structured'}}, 'module_output_name': 'token_embeddings', 'architecture': 'EmbeddingGemma2Model'})
  (1): Pooling({'embedding_dimension': 768, 'pooling_mode': 'mean', 'include_prompt': True})
  (2): Normalize({'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
)

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("Jbbok/my-embeddinggemma-2-sec")
# Run inference
queries = [
    'Why can GE not predict the economic conditions that will exist when a new product is ready to enter service?',
]
documents = [
    'company: GENERAL ELECTRIC CO | ticker: GE | type: 10-K | report_date: 2025-12-31\n**Competitive environment - We are dependent on the maintenance of existing product lines and service relationships, acceptance by our customers of new product and service introductions, competitive pricing and other terms, and technology and innovation leadership for revenue and earnings growth.** The segments in which we, other engine manufacturers and airframers operate, are highly competitive in terms of pricing, product and service quality, product development and entry into service, product durability, customer service, financing terms, the ability to respond to shifts in market demand and the ability to attract and retain skilled talent at all levels, including senior leadership. Our long-term operating results and competitive position depend substantially upon our ability to continue to improve or upgrade current products and services, to have our engines power leading existing and new aircraft platforms, to cultivate, maintain and grow long-term customer relationships and to increase our productivity over time as we perform on long-term service agreements. Our competitive position and success also depend on our ability to develop, introduce, and market new and innovative technology, products, services and platforms, such as the RISE program suite of technologies. The research and development cycle involved in bringing new products to market is often lengthy, it is inherently difficult to predict the economic conditions or competitive dynamics that will exist when any new product is complete, and our investments, to the extent they result in bringing a product to market, may generate weaker returns than we anticipated at the outset. Our capacity to invest in research and development efforts to advance our technologies, products and services, or to otherwise invest in emerging technologies or partner with other industry players, also depends on the financial resources that we have available for such investment relative to other capital allocation priorities. Under-investment in research and development and innovation, or investment in technologies that prove to be less competitive in the future (at the expense of alternative investment opportunities not pursued), could lead to loss of sales of our products or services in the future due to the long product development cycles in our business. The amounts that we do invest in research and development efforts may not lead to the development of new technologies or products on a timely basis or meet the needs of our customers as fully as competitive offerings, and we may face impairment charges for contract fulfillment costs that are capitalized as nonrecurring engineering costs if we determine recovery of the costs is not probable (see Note 1).',
    'company: AMERICAN EXPRESS CO | ticker: AXP | type: 10-K | report_date: 2025-12-31\nWe are also subject to certain privacy, data protection, data management, AI, resiliency, information security and cybersecurity laws in other countries in which we operate, some of which are more stringent and/or expansive than those in the United States and may conflict with each other. The EU and UK General Data Protection Regulations (GDPR) impose legal and compliance obligations on companies that process personal data of individuals in the EU and UK, irrespective of the geographical location of the company, with the potential for significant fines for non-compliance (up to 4 percent of total annual worldwide revenue). The EU and UK GDPR also include requirements concerning the cross-border transfer of personal data and prompt notification of data breaches, in certain circumstances, to affected individuals and supervisory authorities. We are also subject to certain data protection laws in Member States in the EU, which may be more stringent than the EU GDPR. Other countries have also adopted or are considering similar omnibus privacy laws, including Australia, Brazil, Canada, China, India, Japan, the Philippines, Singapore, South Korea and Thailand. Certain countries also require in-country data processing and/or in-country storage of data or for us to provide foreign governments and other third parties broader access to our data and intellectual property. Data breach and operational outage notification laws or regulatory activities to encourage such notifications and regulatory activity and laws around resiliency, business continuity and third-party risk management are also becoming more prevalent in jurisdictions outside the United States in which we operate.\nOur privacy and data protection programs have become the subject of heightened scrutiny and review in certain jurisdictions, including in the EU, and we continue to enhance our privacy program to comply with applicable requirements and regulatory expectations. Our compliance with the various and often diverging legal frameworks around privacy, data protection, AI, resiliency, information security and cybersecurity, as well as increased regulatory and legislative activity in these areas, may result in higher technology, administrative and other operational costs and hinder our ability to deploy and scale technology, innovate quickly and effectively utilize data.\n**Anti-Money Laundering, Countering the Financing of Terrorism, Economic Sanctions and Anti-Corruption Compliance**',
    'company: AMERICAN EXPRESS CO | ticker: AXP | type: 10-K | report_date: 2025-12-31\nAmerican Express cards could become less desirable to consumers and businesses generally due to surcharging, steering or other forms of discrimination, which could result in a decrease in cards-in-force, coverage and transaction volumes, including as a result of related actions we may take to enforce our merchant contractual provisions such as terminating merchant contracts. The impact could vary depending on such factors as: the industry or manner in which a surcharge is levied; how Card Members are surcharged or steered to other card products or payment forms at the point of sale; the ease and speed of implementation for merchants, merchant acquirers, processors, payment facilitators or other merchant service providers, including as a result of new or emerging technologies such as AI and agentic commerce; the size and recurrence of the underlying charges; and whether and to what extent these actions are applied to other forms of payment, including whether it varies depending on the type of card (e.g., credit or debit), product, network, acquirer or issuer. We also increasingly rely on merchant acquirers, processors and payment facilitators to manage certain aspects of our merchant relationships and promote and support the acceptance and usage of our cards, but they may have business interests, strategies or goals that are inconsistent with ours. Discrimination against American Express cards could have a material adverse effect on our business, financial condition and results of operations, particularly where it only or disproportionately impacts credit card usage or card usage generally, our Card Members or our business.',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 768] [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.4498, 0.0411, 0.1949]])

Training Details

Training Dataset

Unnamed Dataset

Size: 3,867 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 100 samples:

	anchor	positive	negative
type	string	string	string
modality	text	text	text
details	min: 10 tokens mean: 24.23 tokens max: 29 tokens	min: 191 tokens mean: 1319.94 tokens max: 4096 tokens	min: 148 tokens mean: 668.62 tokens max: 4096 tokens

Samples:

anchor	positive	negative
`What were NVIDIA's total assets as of January 25, 2026?`	company: NVIDIA CORP \| ticker: NVDA \| type: 10-K \| report_date: 2026-01-25 Consolidated Balance Sheets (In millions, except par value) (columns: Current assets:) ## Assets ## Current assets ### Cash and cash equivalents (us-gaap:CashAndCashEquivalentsAtCarryingValue) - Jan 25, 2026: $10,605 million [=10,605,000,000 USD] - Jan 26, 2025: $8,589 million [=8,589,000,000 USD] ### Marketable securities (nvda:MarketableSecuritiesAndEquitySecuritiesFVNI) - Jan 25, 2026: 51,951 million [=51,951,000,000 USD] - Jan 26, 2025: 34,621 million [=34,621,000,000 USD] ### Accounts receivable, net (us-gaap:AccountsReceivableNetCurrent) - Jan 25, 2026: 38,466 million [=38,466,000,000 USD] - Jan 26, 2025: 23,065 million [=23,065,000,000 USD] ### Inventories (us-gaap:InventoryNet) - Jan 25, 2026: 21,403 million [=21,403,000,000 USD] - Jan 26, 2025: 10,080 million [=10,080,000,000 USD] ### Prepaid expenses and other current assets (us-gaap:PrepaidExpenseAndOtherAssetsCurrent) - Jan 25,...	company: NVIDIA CORP \| ticker: NVDA \| type: 10-K \| report_date: 2026-01-25 Property, equipment and intangible assets acquired but not paid for during fiscal years 2026, 2025, and 2024 were $820 million, $525 million, and $170 million, respectively. Accrued and Other Current Liabilities: (In millions) ### Customer program accruals (nvda:AccruedCustomerProgramsCurrent) - Jan 25, 2026: $5,318 million [=5,318,000,000 USD] - Jan 26, 2025: $4,880 million [=4,880,000,000 USD] ### Accrued purchase consideration (nvda:BusinessCombinationConsiderationPayable) - Jan 25, 2026: 3,921 million [=3,921,000,000 USD] - Jan 26, 2025: 9 million [=9,000,000 USD] ### Product warranty (us-gaap:ProductWarrantyAccrualClassifiedCurrent) - Jan 25, 2026: 2,807 million [=2,807,000,000 USD] - Jan 26, 2025: 1,290 million [=1,290,000,000 USD] ### Excess inventory purchase obligations (1) (nvda:ExcessInventoryPurchaseObligationsCurrent) - Jan 25, 2026: 2,739 million [=2,739,000,000 USD] - Jan 26, 2025: 2,09...
`What were NVIDIA's total assets as of January 25, 2026?`	company: NVIDIA CORP \| ticker: NVDA \| type: 10-K \| report_date: 2026-01-25 Consolidated Balance Sheets (In millions, except par value) (columns: Current assets:) ## Assets ## Current assets ### Cash and cash equivalents (us-gaap:CashAndCashEquivalentsAtCarryingValue) - Jan 25, 2026: $10,605 million [=10,605,000,000 USD] - Jan 26, 2025: $8,589 million [=8,589,000,000 USD] ### Marketable securities (nvda:MarketableSecuritiesAndEquitySecuritiesFVNI) - Jan 25, 2026: 51,951 million [=51,951,000,000 USD] - Jan 26, 2025: 34,621 million [=34,621,000,000 USD] ### Accounts receivable, net (us-gaap:AccountsReceivableNetCurrent) - Jan 25, 2026: 38,466 million [=38,466,000,000 USD] - Jan 26, 2025: 23,065 million [=23,065,000,000 USD] ### Inventories (us-gaap:InventoryNet) - Jan 25, 2026: 21,403 million [=21,403,000,000 USD] - Jan 26, 2025: 10,080 million [=10,080,000,000 USD] ### Prepaid expenses and other current assets (us-gaap:PrepaidExpenseAndOtherAssetsCurrent) - Jan 25,...	company: NVIDIA CORP \| ticker: NVDA \| type: 10-K \| report_date: 2026-01-25 Certain balance sheet components are as follows: Inventories: (In millions) ### Raw materials (us-gaap:InventoryRawMaterialsNetOfReserves) - Jan 25, 2026: $3,807 million [=3,807,000,000 USD] - Jan 26, 2025: $3,408 million [=3,408,000,000 USD] ### Work in process (us-gaap:InventoryWorkInProcessNetOfReserves) - Jan 25, 2026: 8,822 million [=8,822,000,000 USD] - Jan 26, 2025: 3,399 million [=3,399,000,000 USD] ### Finished goods (us-gaap:InventoryFinishedGoodsNetOfReserves) - Jan 25, 2026: 8,774 million [=8,774,000,000 USD] - Jan 26, 2025: 3,273 million [=3,273,000,000 USD] ### Total inventories (1) (us-gaap:InventoryNet) - Jan 25, 2026: $21,403 million [=21,403,000,000 USD] - Jan 26, 2025: $10,080 million [=10,080,000,000 USD]
`What were NVIDIA's total assets as of January 25, 2026?`	company: NVIDIA CORP \| ticker: NVDA \| type: 10-K \| report_date: 2026-01-25 Consolidated Balance Sheets (In millions, except par value) (columns: Current assets:) ## Assets ## Current assets ### Cash and cash equivalents (us-gaap:CashAndCashEquivalentsAtCarryingValue) - Jan 25, 2026: $10,605 million [=10,605,000,000 USD] - Jan 26, 2025: $8,589 million [=8,589,000,000 USD] ### Marketable securities (nvda:MarketableSecuritiesAndEquitySecuritiesFVNI) - Jan 25, 2026: 51,951 million [=51,951,000,000 USD] - Jan 26, 2025: 34,621 million [=34,621,000,000 USD] ### Accounts receivable, net (us-gaap:AccountsReceivableNetCurrent) - Jan 25, 2026: 38,466 million [=38,466,000,000 USD] - Jan 26, 2025: 23,065 million [=23,065,000,000 USD] ### Inventories (us-gaap:InventoryNet) - Jan 25, 2026: 21,403 million [=21,403,000,000 USD] - Jan 26, 2025: 10,080 million [=10,080,000,000 USD] ### Prepaid expenses and other current assets (us-gaap:PrepaidExpenseAndOtherAssetsCurrent) - Jan 25,...	company: NVIDIA CORP \| ticker: NVDA \| type: 10-K \| report_date: 2026-01-25 Consolidated Statements of Cash Flows (In millions) ## Cash flows from operating activities ### Net income (Year Ended) (us-gaap:NetIncomeLoss) - Jan 25, 2026: $120,067 million [=120,067,000,000 USD] - Jan 26, 2025: $72,880 million [=72,880,000,000 USD] - Jan 28, 2024: $29,760 million [=29,760,000,000 USD] ## Adjustments to reconcile net income to net cash provided by operating activities ### Stock-based compensation expense (Year Ended) (us-gaap:ShareBasedCompensation) - Jan 25, 2026: 6,386 million [=6,386,000,000 USD] - Jan 26, 2025: 4,737 million [=4,737,000,000 USD] - Jan 28, 2024: 3,549 million [=3,549,000,000 USD] ### Depreciation and amortization (Year Ended) (us-gaap:DepreciationDepletionAndAmortization) - Jan 25, 2026: 2,843 million [=2,843,000,000 USD] - Jan 26, 2025: 1,864 million [=1,864,000,000 USD] - Jan 28, 2024: 1,508 million [=1,508,000,000 USD] ### Gains on non-marketable equit...

Loss: CachedMultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "mini_batch_size": 16,
    "mini_batch_num_tokens": null,
    "gather_across_devices": false,
    "directions": [
        "query_to_doc"
    ],
    "partition_mode": "joint",
    "hardness_mode": null,
    "hardness_strength": 0.0
}

Evaluation Dataset

Unnamed Dataset

Size: 480 evaluation samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 100 samples:

	anchor	positive	negative
type	string	string	string
modality	text	text	text
details	min: 12 tokens mean: 22.23 tokens max: 45 tokens	min: 120 tokens mean: 1210.79 tokens max: 4096 tokens	min: 100 tokens mean: 1068.83 tokens max: 4096 tokens

Samples:

anchor	positive	negative
`Why can GE not predict the economic conditions that will exist when a new product is ready to enter service?`	company: GENERAL ELECTRIC CO \| ticker: GE \| type: 10-K \| report_date: 2025-12-31 Competitive environment - We are dependent on the maintenance of existing product lines and service relationships, acceptance by our customers of new product and service introductions, competitive pricing and other terms, and technology and innovation leadership for revenue and earnings growth. The segments in which we, other engine manufacturers and airframers operate, are highly competitive in terms of pricing, product and service quality, product development and entry into service, product durability, customer service, financing terms, the ability to respond to shifts in market demand and the ability to attract and retain skilled talent at all levels, including senior leadership. Our long-term operating results and competitive position depend substantially upon our ability to continue to improve or upgrade current products and services, to have our engines power leading existing and new aircraft pla...	company: GENERAL ELECTRIC CO \| ticker: GE \| type: 10-K \| report_date: 2025-12-31 Customers and counterparties - Global economic, industry-specific or other developments that weaken the financial condition, soundness or continuity of significant customers, governments, government programs or other parties we deal with can adversely affect our business, results of operations and cash flows. Our business and operating results have been, and will continue to be, affected by worldwide economic conditions, including conditions in the aerospace and defense sector. Activity in our sector is also particularly influenced by the actions of a small group of large original equipment manufacturers, as well as large airlines in various geographies. We have significant business with, and credit exposure to, some of our largest customers and accordingly our performance can be adversely affected by challenges that individual customers or the industry faces related to factors such as competition, reg...
`What privacy and data protection risks does American Express face, and how could they affect the company?`	company: AMERICAN EXPRESS CO \| ticker: AXP \| type: 10-K \| report_date: 2025-12-31 We are also subject to certain privacy, data protection, data management, AI, resiliency, information security and cybersecurity laws in other countries in which we operate, some of which are more stringent and/or expansive than those in the United States and may conflict with each other. The EU and UK General Data Protection Regulations (GDPR) impose legal and compliance obligations on companies that process personal data of individuals in the EU and UK, irrespective of the geographical location of the company, with the potential for significant fines for non-compliance (up to 4 percent of total annual worldwide revenue). The EU and UK GDPR also include requirements concerning the cross-border transfer of personal data and prompt notification of data breaches, in certain circumstances, to affected individuals and supervisory authorities. We are also subject to certain data protection laws in Member State...	company: AMERICAN EXPRESS CO \| ticker: AXP \| type: 10-K \| report_date: 2025-12-31 If we are not able to protect our intellectual property rights, or successfully defend against any infringement or misappropriation assertions brought against us, our revenue and profitability could be negatively affected. We rely on a variety of measures to protect our intellectual property rights and control access to, and distribution of, our trade secrets and other proprietary information. These measures may not prevent infringement of our intellectual property rights or misappropriation of our proprietary information and a resulting loss of competitive advantage. Our ability to detect infringements of our intellectual property, enforce intellectual property rights and prevent disclosure of our trade secrets and other proprietary information may be limited and such efforts may be costly. In addition, competitors or other third parties may allege that our products, systems, processes or technolog...
`How could merchant surcharging or steering American Express cards to other payment options affect American Express?`	company: AMERICAN EXPRESS CO \| ticker: AXP \| type: 10-K \| report_date: 2025-12-31 American Express cards could become less desirable to consumers and businesses generally due to surcharging, steering or other forms of discrimination, which could result in a decrease in cards-in-force, coverage and transaction volumes, including as a result of related actions we may take to enforce our merchant contractual provisions such as terminating merchant contracts. The impact could vary depending on such factors as: the industry or manner in which a surcharge is levied; how Card Members are surcharged or steered to other card products or payment forms at the point of sale; the ease and speed of implementation for merchants, merchant acquirers, processors, payment facilitators or other merchant service providers, including as a result of new or emerging technologies such as AI and agentic commerce; the size and recurrence of the underlying charges; and whether and to what extent these actions are...	company: AMERICAN EXPRESS CO \| ticker: AXP \| type: 10-K \| report_date: 2025-12-31 Merchant Acquiring Business Our GMNS reportable operating segment builds and manages relationships with millions of merchants around the world that choose to accept American Express cards. This includes signing new merchants to accept our cards, agreeing on the discount rate (a fee charged to the merchant for accepting our cards) and handling servicing for merchants. We also build and maintain relationships with merchant acquirers, processors and payment facilitators to manage aspects of our merchant services business. For example, through our OptBlue® merchant-acquiring program, these third parties contract directly with small merchants for card acceptance on our network and determine merchant pricing. We continue to grow merchant acceptance of American Express cards around the world and work with merchant partners so that our Card Members are warmly welcomed and encouraged to spend in the millions of ...

Loss: CachedMultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "mini_batch_size": 16,
    "mini_batch_num_tokens": null,
    "gather_across_devices": false,
    "directions": [
        "query_to_doc"
    ],
    "partition_mode": "joint",
    "hardness_mode": null,
    "hardness_strength": 0.0
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
learning_rate: 2e-05
warmup_steps: 0.1
bf16: True
per_device_eval_batch_size: 32
load_best_model_at_end: True
seed: 0
prompts: {'anchor': 'task: search result | query: ', 'positive': 'title: none | text: ', 'negative': 'title: none | text: '}
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

per_device_train_batch_size: 32
num_train_epochs: 3
max_steps: -1
learning_rate: 2e-05
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_steps: 0.1
optim: adamw_torch_fused
optim_args: None
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
optim_target_modules: None
gradient_accumulation_steps: 1
average_tokens_across_devices: True
max_grad_norm: 1.0
label_smoothing_factor: 0.0
bf16: True
fp16: False
bf16_full_eval: False
fp16_full_eval: False
tf32: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
use_liger_kernel: False
liger_kernel_config: None
use_cache: False
neftune_noise_alpha: None
torch_empty_cache_steps: None
auto_find_batch_size: False
log_on_each_node: True
logging_nan_inf_filter: True
include_num_input_tokens_seen: no
log_level: passive
log_level_replica: warning
disable_tqdm: False
project: huggingface
trackio_space_id: None
trackio_bucket_id: None
trackio_static_space_id: None
per_device_eval_batch_size: 32
prediction_loss_only: True
eval_on_start: False
eval_do_concat_batches: True
eval_use_gather_object: False
eval_accumulation_steps: None
include_for_metrics: []
batch_eval_metrics: False
save_only_model: False
save_on_each_node: False
enable_jit_checkpoint: False
push_to_hub: False
hub_private_repo: None
hub_model_id: None
hub_strategy: every_save
hub_always_push: False
hub_revision: None
load_best_model_at_end: True
ignore_data_skip: False
restore_callback_states_from_checkpoint: False
full_determinism: False
seed: 0
data_seed: None
use_cpu: False
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
parallelism_config: None
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_pin_memory: True
dataloader_persistent_workers: False
dataloader_prefetch_factor: None
dataloader_multiprocessing_context: None
dataloader_in_order: True
remove_unused_columns: True
label_names: None
train_sampling_strategy: random
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
ddp_static_graph: None
ddp_backend: None
ddp_timeout: 1800
fsdp: None
fsdp_config: None
deepspeed: None
debug: []
skip_memory_metrics: True
do_predict: False
resume_from_checkpoint: None
local_rank: -1
prompts: {'anchor': 'task: search result | query: ', 'positive': 'title: none | text: ', 'negative': 'title: none | text: '}
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}
warmup_ratio: None

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.2479	30	1.0390	-
0.4959	60	0.3680	-
0.7438	90	0.2735	-
0.9917	120	0.2537	-
1.0	121	-	0.6994
1.2397	150	0.1704	-
1.4876	180	0.1617	-
1.7355	210	0.1602	-
1.9835	240	0.1513	-
2.0	242	-	0.6717
2.2314	270	0.1670	-
2.4793	300	0.1111	-
2.7273	330	0.1506	-
2.9752	360	0.1101	-
3.0	363	-	0.6962

The bold row denotes the saved checkpoint.

Training Time

Training: 2.7 hours

Framework Versions

Python: 3.13.16
Sentence Transformers: 6.2.0
Transformers: 5.22.1
PyTorch: 2.11.0+cu130
Accelerate: 1.15.0
Datasets: 5.1.0
Tokenizers: 0.23.3

Additional Resources

Training and Finetuning Embedding Models with Sentence Transformers: the end-to-end guide for training or finetuning Sentence Transformer models.
Introduction to Matryoshka Embedding Models: variable-size embeddings that can be truncated with minimal quality loss.
Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval: post-training compression of embedding vectors.
Multimodal Embedding & Reranker Models with Sentence Transformers: use text, image, audio, and video models through the same API.
Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers: train multimodal embedding models, with a Visual Document Retrieval walkthrough.

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",
}

CachedMultipleNegativesRankingLoss

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}

Downloads last month: 34

Safetensors

Model size

0.3B params

Tensor type

F32

Model tree for Jbbok/my-embeddinggemma-2-sec

Base model

google/embeddinggemma-2

Finetuned

(164)

this model

Papers for Jbbok/my-embeddinggemma-2-sec