This is a sentence-transformers model finetuned from google/embeddinggemma-300m. 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: google/embeddinggemma-300m
Maximum Sequence Length: 2048 tokens
Output Dimensionality: 768 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': 2048, 'do_lower_case': False, 'architecture': 'Gemma3TextModel'})
  (1): Pooling({'word_embedding_dimension': 768, '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': 768, 'out_features': 3072, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
  (3): Dense({'in_features': 3072, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
  (4): 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("Maksim-KOS/embeddinggemma-300m-saturn-planet-v2")
# Run inference
queries = [
    "\u0428\u043a\u0443\u0440\u043a\u0430 \u0432\u043e\u0434\u043e\u0441\u0442\u043e\u0439\u043a\u0430\u044f \u043d\u0430 \u0442\u043a\u0430\u043d\u0435\u0432\u043e\u0439 \u043e\u0441\u043d\u043e\u0432\u0435 P80 80\u043c\u043c \u0440\u0443\u043b\u043e\u043d 20\u043c",
]
documents = [
    'Шкурка шлифовальная водостойкая на тканевой основе 805 мм, 20Н (P80), рулон 20 м',
    'Шкурка шлифовальная водостойкая на тканевой основе 805 мм 25Н (P60) 1 п.м.',
    'Перчатки КЩС тип 2 тонкие',
]
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([[ 0.8399,  0.2156, -0.0400]])

Evaluation

Metrics

Triplet

Dataset: hard-neg-eval
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9931

Training Details

Training Dataset

Unnamed Dataset

Size: 124,228 training samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 8 tokens
mean: 21.2 tokens
max: 58 tokens
min: 10 tokens
mean: 26.83 tokens
max: 52 tokens
min: 7 tokens
mean: 27.71 tokens
max: 51 tokens

Samples:

anchor	positive	negative
`Кабель Камкабель КГтп 2х2,5 220 В (100 м), шт`	`Кабель КГтп-ХЛ 2х2,5мм2, гибкий, черный (бухта-100п.м.)`	`Кабель КГтп-ХЛ 2х1,5мм2 TDM гибкий, черный (бухта-100п.м.)`
`Клей для плитки Основит Стройбриг С0T АС12 25 кг`	`Клей для плитки Стройбриг Гранификс АС12 усиленный (25 кг)`	`Клей для плитки Основит Мастпликс AC121 Т выравнивающий (25 кг)`
`линолеум juteks аванта 4м`	`Линолеум бытовой Juteks Avanta NXT Kortes 4 (4 м)`	`Линолеум бытовой Juteks Avanta Fort 5 (4 м)`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false 5}

Evaluation Dataset

Unnamed Dataset

Size: 13,804 evaluation samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 7 tokens
mean: 21.33 tokens
max: 59 tokens
min: 7 tokens
mean: 26.59 tokens
max: 54 tokens
min: 8 tokens
mean: 27.72 tokens
max: 55 tokens

Samples:

anchor	positive	negative
`Свая винтовая d76х2000 мм, шт`	`Свая винтовая d=76 мм, L=2000 мм, d винтовой части = 220 мм`	`Столб заборный винт., L=3000 мм (60х60 мм L=2000 мм, d=51 мм L=1000 мм)`
`Нивелир лазерный Condtrol Neo Х200 (1-2-115) с магнитной шайбой, шт`	`Лазерный нивелир Condtrol NEO X200`	`Нивелир лазерный Condtrol NEO X2-360`
`керамогранит светло-коричневый 594х147 мм`	`Керамогранит Березакерамика Шеми GP, светло-коричневый, 594х147х9 мм`	`Керамогранит Керамин Портланд 3, мат., светло-бежевый, 600х600х10 мм`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false 5}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
gradient_accumulation_steps: 8
learning_rate: 1e-05
weight_decay: 0.01
num_train_epochs: 10
lr_scheduler_type: cosine
warmup_ratio: 0.1
load_best_model_at_end: True
optim: adamw_torch
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: 32
per_device_eval_batch_size: 32
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 8
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 1e-05
weight_decay: 0.01
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 10
max_steps: -1
lr_scheduler_type: cosine
lr_scheduler_kwargs: None
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
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: 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}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
project: huggingface
trackio_space_id: trackio
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
hub_revision: None
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: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	hard-neg-eval_cosine_accuracy
0.1030	50	0.2049	-	-
0.2060	100	0.0909	0.0862	0.9690
0.3090	150	0.0768	-	-
0.4121	200	0.0652	0.0626	0.9769
0.5151	250	0.0597	-	-
0.6181	300	0.0592	0.0564	0.9793
0.7211	350	0.0572	-	-
0.8241	400	0.0484	0.0497	0.9814
0.9271	450	0.0492	-	-
1.0288	500	0.0444	0.0503	0.9811
1.1319	550	0.039	-	-
1.2349	600	0.0399	0.0448	0.9830
1.3379	650	0.0397	-	-
1.4409	700	0.0351	0.0405	0.9851
1.5439	750	0.0368	-	-
1.6469	800	0.0389	0.0375	0.9851
1.7499	850	0.0353	-	-
1.8529	900	0.0366	0.0354	0.9859
1.9560	950	0.0314	-	-
2.0577	1000	0.0264	0.0345	0.9865
2.1607	1050	0.0234	-	-
2.2637	1100	0.0245	0.0343	0.9869
2.3667	1150	0.0229	-	-
2.4697	1200	0.0235	0.0335	0.9875
2.5728	1250	0.0226	-	-
2.6758	1300	0.0249	0.0293	0.9888
2.7788	1350	0.0224	-	-
2.8818	1400	0.023	0.0278	0.9904
2.9848	1450	0.0233	-	-
3.0865	1500	0.0156	0.0275	0.9896
3.1895	1550	0.0145	-	-
3.2926	1600	0.0142	0.0269	0.9896
3.3956	1650	0.015	-	-
3.4986	1700	0.017	0.0258	0.9897
3.6016	1750	0.0152	-	-
3.7046	1800	0.0148	0.0246	0.9905
3.8076	1850	0.0149	-	-
3.9106	1900	0.0166	0.0251	0.9902
4.0124	1950	0.0173	-	-
4.1154	2000	0.0102	0.0238	0.9907
4.2184	2050	0.0092	-	-
4.3214	2100	0.0113	0.0221	0.9920
4.4244	2150	0.0082	-	-
4.5274	2200	0.0091	0.0221	0.9921
4.6304	2250	0.0107	-	-
4.7335	2300	0.0111	0.0208	0.9928
4.8365	2350	0.0086	-	-
4.9395	2400	0.0103	0.0202	0.9921
5.0412	2450	0.0076	-	-
5.1442	2500	0.0066	0.0193	0.9933
5.2472	2550	0.0068	-	-
5.3502	2600	0.006	0.0198	0.9932
5.4533	2650	0.0064	-	-
5.5563	2700	0.0057	0.0190	0.9931
5.6593	2750	0.0057	-	-
5.7623	2800	0.0061	0.0184	0.9931

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.12.3
Sentence Transformers: 5.2.0
Transformers: 4.57.6
PyTorch: 2.9.1+cu128
Accelerate: 1.12.0
Datasets: 4.5.0
Tokenizers: 0.22.2

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",
}

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}
}

	anchor	positive	negative
type	string	string	string
details	min: 8 tokens mean: 21.2 tokens max: 58 tokens	min: 10 tokens mean: 26.83 tokens max: 52 tokens	min: 7 tokens mean: 27.71 tokens max: 51 tokens