You can deploy this model and start earning money today!

Views

No views yet

Google SigLIP2 (512x512 resolution) model trained on COCO Captions

Name: google-siglip2-base-512-coco
Availability: InStock

This is a sentence-transformers model finetuned from google/siglip2-base-patch16-512 on the coco_captions dataset. It maps sentences & paragraphs to a None-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/siglip2-base-patch16-512
Maximum Sequence Length: None tokens
Output Dimensionality: None dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- coco_captions
Language: en
License: apache-2.0

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': 'get_text_features', 'method_output_name': None}, 'image': {'method': 'get_image_features', 'method_output_name': None}}, 'module_output_name': 'sentence_embedding', 'architecture': 'SiglipModel'})
)

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("tomaarsen/google-siglip2-base-512-coco")
# Run inference
sentences = [
    'A large desk by a window is neatly arranged.',
    'A long hot dog on a plate on a table.',
    'A lady sitting at an enormous dining table with lots of food.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.1125, 0.2752],
#         [0.1125, 1.0000, 0.3446],
#         [0.2752, 0.3446, 1.0000]])

Evaluation

Metrics

Information Retrieval

Datasets: coco-eval and coco-test
Evaluated with InformationRetrievalEvaluator

Metric	coco-eval	coco-test
cosine_accuracy@1	0.627	0.65
cosine_accuracy@3	0.892	0.874
cosine_accuracy@5	0.948	0.939
cosine_accuracy@10	0.98	0.98
cosine_precision@1	0.627	0.65
cosine_precision@3	0.2973	0.2913
cosine_precision@5	0.1896	0.1878
cosine_precision@10	0.098	0.098
cosine_recall@1	0.627	0.65
cosine_recall@3	0.892	0.874
cosine_recall@5	0.948	0.939
cosine_recall@10	0.98	0.98
cosine_ndcg@10	0.8152	0.8236
cosine_mrr@10	0.7606	0.7723
cosine_map@100	0.7614	0.7731

Training Details

Training Dataset

coco_captions

Dataset: coco_captions at a2ed90d
Size: 10,000 training samples
Columns: image and caption
Approximate statistics based on the first 1000 samples:
image caption
type PIL.JpegImagePlugin.JpegImageFile string
details
min: 28 characters
mean: 52.56 characters
max: 156 characters

Samples:

image	caption
`<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=640x480 at 0x20DAA211010>`	`A woman wearing a net on her head cutting a cake.`
`<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=640x480 at 0x226133D1850>`	`A woman cutting a large white sheet cake.`
`<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=640x480 at 0x226133D2050>`	`A woman wearing a hair net cutting a large sheet cake.`

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

Evaluation Dataset

coco_captions

Dataset: coco_captions at a2ed90d
Size: 1,000 evaluation samples
Columns: image and caption
Approximate statistics based on the first 1000 samples:
image caption
type PIL.JpegImagePlugin.JpegImageFile string
details
min: 27 characters
mean: 52.45 characters
max: 151 characters

Samples:

image	caption
`<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=500x336 at 0x2261341DF50>`	`A child holding a flowered umbrella and petting a yak.`
`<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=500x336 at 0x20D01377750>`	`A young man holding an umbrella next to a herd of cattle.`
`<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=500x336 at 0x226133D2110>`	`a young boy barefoot holding an umbrella touching the horn of a cow`

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: 16
per_device_eval_batch_size: 16
learning_rate: 2e-05
num_train_epochs: 1
warmup_ratio: 0.1
bf16: True
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: 16
per_device_eval_batch_size: 16
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 2e-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: 0.1
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
use_cpu: False
seed: 42
data_seed: None
jit_mode_eval: False
bf16: True
fp16: False
half_precision_backend: None
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 0
ddp_backend: None
tpu_num_cores: None
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_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': 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
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
optim_args: None
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
hub_revision: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_for_metrics: []
eval_do_concat_batches: True
mp_parameters:
auto_find_batch_size: False
full_determinism: False
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	coco-eval_cosine_ndcg@10	coco-test_cosine_ndcg@10
-1	-1	-	-	0.1900	-
0.0112	7	2.4513	-	-	-
0.0224	14	2.4538	-	-	-
0.0336	21	2.4508	-	-	-
0.0448	28	2.3998	-	-	-
0.056	35	2.3021	-	-	-
0.0672	42	2.2435	-	-	-
0.0784	49	1.5821	-	-	-
0.0896	56	1.5381	-	-	-
0.1008	63	1.652	1.4376	0.3890	-
0.112	70	1.175	-	-	-
0.1232	77	1.2468	-	-	-
0.1344	84	0.9952	-	-	-
0.1456	91	0.9492	-	-	-
0.1568	98	0.9434	-	-	-
0.168	105	0.8598	-	-	-
0.1792	112	0.6562	-	-	-
0.1904	119	0.8115	-	-	-
0.2016	126	0.5666	0.6434	0.6437	-
0.2128	133	0.4335	-	-	-
0.224	140	0.3687	-	-	-
0.2352	147	0.52	-	-	-
0.2464	154	0.5098	-	-	-
0.2576	161	0.5707	-	-	-
0.2688	168	0.5349	-	-	-
0.28	175	0.5768	-	-	-
0.2912	182	0.3053	-	-	-
0.3024	189	0.293	0.4603	0.6986	-
0.3136	196	0.4413	-	-	-
0.3248	203	0.4706	-	-	-
0.336	210	0.3618	-	-	-
0.3472	217	0.4465	-	-	-
0.3584	224	0.3968	-	-	-
0.3696	231	0.3413	-	-	-
0.3808	238	0.3039	-	-	-
0.392	245	0.2978	-	-	-
0.4032	252	0.4956	0.3713	0.7410	-
0.4144	259	0.3441	-	-	-
0.4256	266	0.4604	-	-	-
0.4368	273	0.4723	-	-	-
0.448	280	0.2957	-	-	-
0.4592	287	0.262	-	-	-
0.4704	294	0.2205	-	-	-
0.4816	301	0.2546	-	-	-
0.4928	308	0.3489	-	-	-
0.504	315	0.2028	0.3304	0.7648	-
0.5152	322	0.2659	-	-	-
0.5264	329	0.356	-	-	-
0.5376	336	0.3322	-	-	-
0.5488	343	0.1783	-	-	-
0.56	350	0.3221	-	-	-
0.5712	357	0.2213	-	-	-
0.5824	364	0.303	-	-	-
0.5936	371	0.2349	-	-	-
0.6048	378	0.309	0.2987	0.7756	-
0.616	385	0.2494	-	-	-
0.6272	392	0.1605	-	-	-
0.6384	399	0.21	-	-	-
0.6496	406	0.1258	-	-	-
0.6608	413	0.2092	-	-	-
0.672	420	0.2701	-	-	-
0.6832	427	0.181	-	-	-
0.6944	434	0.2653	-	-	-
0.7056	441	0.3197	0.2817	0.7883	-
0.7168	448	0.2991	-	-	-
0.728	455	0.175	-	-	-
0.7392	462	0.159	-	-	-
0.7504	469	0.1689	-	-	-
0.7616	476	0.2212	-	-	-
0.7728	483	0.1601	-	-	-
0.784	490	0.1509	-	-	-
0.7952	497	0.2051	-	-	-
0.8064	504	0.2972	0.2615	0.8024	-
0.8176	511	0.1601	-	-	-
0.8288	518	0.2681	-	-	-
0.84	525	0.1635	-	-	-
0.8512	532	0.1868	-	-	-
0.8624	539	0.1284	-	-	-
0.8736	546	0.1414	-	-	-
0.8848	553	0.1474	-	-	-
0.896	560	0.1482	-	-	-
0.9072	567	0.1718	0.2459	0.8152	-
0.9184	574	0.1705	-	-	-
0.9296	581	0.1761	-	-	-
0.9408	588	0.232	-	-	-
0.952	595	0.1619	-	-	-
0.9632	602	0.1088	-	-	-
0.9744	609	0.1874	-	-	-
0.9856	616	0.1502	-	-	-
0.9968	623	0.22	-	-	-
-1	-1	-	-	-	0.8236

Environmental Impact

Carbon emissions were measured using CodeCarbon.

Energy Consumed: 0.056 kWh
Carbon Emitted: 0.015 kg of CO2
Hours Used: 0.175 hours

Training Hardware

On Cloud: No
GPU Model: 1 x NVIDIA GeForce RTX 3090
CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
RAM Size: 31.78 GB

Framework Versions

Python: 3.11.6
Sentence Transformers: 5.2.0.dev0
Transformers: 4.57.0.dev0
PyTorch: 2.8.0+cu128
Accelerate: 1.6.0
Datasets: 3.6.0
Tokenizers: 0.22.1

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",
}

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}
}

	image	caption
type	PIL.JpegImagePlugin.JpegImageFile	string
details		min: 28 characters mean: 52.56 characters max: 156 characters