Skip to content

fix: Add WebSSL models #2604

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
May 1, 2025
Merged

fix: Add WebSSL models #2604

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
2892a82
add 2 web SSL dino models
isaac-chung Apr 27, 2025
dd47d01
add models from collection and revisions
isaac-chung Apr 30, 2025
fa437d1
update memory_usage_mb and embed dim
isaac-chung Apr 30, 2025
44ef88b
use automodel instead
isaac-chung Apr 30, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
379 changes: 379 additions & 0 deletions mteb/models/dino_models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -225,3 +225,382 @@ def get_fused_embeddings(
use_instructions=False,
training_datasets=dinov2_training_datasets,
)

webssl_dino_training_datasets = {
# MetaCLIP 2B samples
}

webssl_dino300m_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino300m-full2b-224",
),
name="facebook/webssl-dino300m-full2b-224",
languages=["eng-Latn"],
revision="8529cdb3fb75014932af3b896455fc21c386168e",
release_date="2025-04-24",
modalities=["image"],
n_parameters=304_000_000,
memory_usage_mb=1158,
max_tokens=None,
embed_dim=1024,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino300m-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino1b_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino1b-full2b-224",
),
name="facebook/webssl-dino1b-full2b-224",
languages=["eng-Latn"],
revision="d3bf033d9c8cc62ea9e73c40956642cad2ec568a",
release_date="2025-04-24",
modalities=["image"],
n_parameters=1_130_000_000,
memory_usage_mb=4329,
max_tokens=None,
embed_dim=1536,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino1b-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino2b_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino2b-full2b-224",
),
name="facebook/webssl-dino2b-full2b-224",
languages=["eng-Latn"],
revision="cd5893e3fd2e988eb716792049b3dd53b3f1b68b",
release_date="2025-04-24",
modalities=["image"],
n_parameters=2_080_000_000,
memory_usage_mb=7951,
max_tokens=None,
embed_dim=2688,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino2b-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino3b_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino3b-full2b-224",
),
name="facebook/webssl-dino3b-full2b-224",
languages=["eng-Latn"],
revision="2d015c340b16bc47bc6557fcb4e6c83a9d4aa1d3",
release_date="2025-04-24",
modalities=["image"],
n_parameters=3_000_000_000,
memory_usage_mb=11247,
max_tokens=None,
embed_dim=3072,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino3b-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino5b_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino5b-full2b-224",
),
name="facebook/webssl-dino5b-full2b-224",
languages=["eng-Latn"],
revision="88006b18b9af369f6c611db7a64d908bde3714e0",
release_date="2025-04-24",
modalities=["image"],
n_parameters=5_000_000_000,
memory_usage_mb=18838,
max_tokens=None,
embed_dim=3584,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino5b-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino7b_full8b_224 = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino7b-full8b-224",
),
name="facebook/webssl-dino7b-full8b-224",
languages=["eng-Latn"],
revision="c6085463ea680043042a80c6d41db2c65e85f466",
release_date="2025-04-24",
modalities=["image"],
n_parameters=7_000_000_000,
memory_usage_mb=24605,
max_tokens=None,
embed_dim=4096,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino7b-full8b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino7b_full8b_378 = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino7b-full8b-378",
),
name="facebook/webssl-dino7b-full8b-378",
languages=["eng-Latn"],
revision="53c8c5b43070bd2ddb3f66161140408ce832301f",
release_date="2025-04-24",
modalities=["image"],
n_parameters=7_000_000_000,
memory_usage_mb=24613,
max_tokens=None,
embed_dim=4096,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino7b-full8b-378",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino7b_full8b_518 = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino7b-full8b-518",
),
name="facebook/webssl-dino7b-full8b-518",
languages=["eng-Latn"],
revision="aee350d2c5e3e5fdb7ee6985291d808ea5eef431",
release_date="2025-04-24",
modalities=["image"],
n_parameters=7_000_000_000,
memory_usage_mb=24623,
max_tokens=None,
embed_dim=4096,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino7b-full8b-518",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino2b_light2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino2b-light2b-224",
),
name="facebook/webssl-dino2b-light2b-224",
languages=["eng-Latn"],
revision="633a663f304e63cc3cbec3f7f9ca2fbc94736128",
release_date="2025-04-24",
modalities=["image"],
n_parameters=2_000_000_000,
memory_usage_mb=7951,
max_tokens=None,
embed_dim=2688,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino2b-light2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino2b_heavy2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino2b-heavy2b-224",
),
name="facebook/webssl-dino2b-heavy2b-224",
languages=["eng-Latn"],
revision="9f46eb0c0129656a1ef195fde072e3765abdb7c6",
release_date="2025-04-24",
modalities=["image"],
n_parameters=2_000_000_000,
memory_usage_mb=7951,
max_tokens=None,
embed_dim=2688,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino2b-heavy2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino3b_light2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino3b-light2b-224",
),
name="facebook/webssl-dino3b-light2b-224",
languages=["eng-Latn"],
revision="4d0160f60673805431f4ad14983e712ed88be5b8",
release_date="2025-04-24",
modalities=["image"],
n_parameters=3_000_000_000,
memory_usage_mb=11247,
max_tokens=None,
embed_dim=3072,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino3b-light2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_dino3b_heavy2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-dino3b-heavy2b-224",
),
name="facebook/webssl-dino3b-heavy2b-224",
languages=["eng-Latn"],
revision="dd39c2910747561b332285d96c4dce0bdb240775",
release_date="2025-04-24",
modalities=["image"],
n_parameters=3_000_000_000,
memory_usage_mb=11247,
max_tokens=None,
embed_dim=3072,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-dino3b-heavy2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_mae300m_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-mae300m-full2b-224",
),
name="facebook/webssl-mae300m-full2b-224",
languages=["eng-Latn"],
revision="4655a0ac1726c206ba14d5ccb26758c62a4d03b0",
release_date="2025-04-24",
modalities=["image"],
n_parameters=304_000_000,
memory_usage_mb=1161,
max_tokens=None,
embed_dim=1024,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-mae300m-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_mae700m_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-mae700m-full2b-224",
),
name="facebook/webssl-mae700m-full2b-224",
languages=["eng-Latn"],
revision="c32be382e757d73a178de1ead62c27391d4b4280",
release_date="2025-04-24",
modalities=["image"],
n_parameters=700_000_000,
memory_usage_mb=2412,
max_tokens=None,
embed_dim=1280,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-mae700m-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)

webssl_mae1b_full2b = ModelMeta(
loader=partial(
DINOModelWrapper,
model_name="facebook/webssl-mae1b-full2b-224",
),
name="facebook/webssl-mae1b-full2b-224",
languages=["eng-Latn"],
revision="5880aefedbad8db0f44d27358f6f08e8576f70fc",
release_date="2025-04-24",
modalities=["image"],
n_parameters=1_000_000_000,
memory_usage_mb=4337,
max_tokens=None,
embed_dim=1536,
license="cc-by-nc-4.0",
open_weights=True,
public_training_code="",
public_training_data=None,
framework=["PyTorch"],
reference="https://huggingface.co/facebook/webssl-mae1b-full2b-224",
similarity_fn_name=None,
use_instructions=False,
training_datasets=webssl_dino_training_datasets,
)