mteb 2.7.2__py3-none-any.whl → 2.7.4__py3-none-any.whl

mteb/models/model_implementations/colqwen_models.py

@@ -1,18 +1,23 @@
+from __future__ import annotations
+
 import logging
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import torch
-from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
 from mteb._requires_package import (
     requires_image_dependencies,
     requires_package,
 )
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
 
 from .colpali_models import (
     COLPALI_CITATION,
@@ -219,6 +224,7 @@ colqwen2 = ModelMeta(
     release_date="2025-11-03",
     modalities=["image", "text"],
     n_parameters=2_210_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=7200,
     max_tokens=32768,
     embed_dim=128,
@@ -246,6 +252,7 @@ colqwen2_5 = ModelMeta(
     release_date="2025-01-31",
     modalities=["image", "text"],
     n_parameters=3_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=7200,
     max_tokens=128000,
     embed_dim=128,
@@ -290,6 +297,7 @@ colqwen3_8b = ModelMeta(
     release_date="2025-11-26",
     modalities=["image", "text"],
     n_parameters=8_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=16724,
     max_tokens=262144,
     embed_dim=320,
@@ -314,6 +322,7 @@ colqwen3_4b = ModelMeta(
     release_date="2025-11-26",
     modalities=["image", "text"],
     n_parameters=4_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=8466,
     max_tokens=262144,
     embed_dim=320,
@@ -329,32 +338,6 @@ colqwen3_4b = ModelMeta(
     citation=TOMORO_CITATION,
 )
 
-colnomic_7b = ModelMeta(
-    loader=ColQwen2_5Wrapper,
-    loader_kwargs=dict(
-        torch_dtype=torch.float16,
-    ),
-    name="nomic-ai/colnomic-embed-multimodal-7b",
-    model_type=["late-interaction"],
-    languages=["eng-Latn"],
-    revision="530094e83a40ca4edcb5c9e5ddfa61a4b5ea0d2f",
-    release_date="2025-03-31",
-    modalities=["image", "text"],
-    n_parameters=7_000_000_000,
-    memory_usage_mb=14400,
-    max_tokens=128000,
-    embed_dim=128,
-    license="apache-2.0",
-    open_weights=True,
-    public_training_code="https://github.com/nomic-ai/colpali",
-    public_training_data="https://huggingface.co/datasets/vidore/colpali_train_set",
-    framework=["ColPali", "safetensors"],
-    reference="https://huggingface.co/nomic-ai/colnomic-embed-multimodal-7b",
-    similarity_fn_name="MaxSim",
-    use_instructions=True,
-    training_datasets=COLPALI_TRAINING_DATA,
-    citation=COLPALI_CITATION,
-)
 
 COLNOMIC_CITATION = """
 @misc{nomicembedmultimodal2025,
@@ -386,6 +369,7 @@ colnomic_3b = ModelMeta(
     release_date="2025-03-31",
     modalities=["image", "text"],
     n_parameters=3_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=7200,
     max_tokens=128000,
     embed_dim=128,
@@ -402,7 +386,7 @@ colnomic_3b = ModelMeta(
 )
 
 colnomic_7b = ModelMeta(
-    loader=ColQwen2Wrapper,
+    loader=ColQwen2_5Wrapper,
     loader_kwargs=dict(
         torch_dtype=torch.float16,
     ),
@@ -451,6 +435,7 @@ evoqwen25_vl_retriever_3b_v1 = ModelMeta(
     release_date="2025-11-04",
     modalities=["image", "text"],
     n_parameters=3_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=7200,
     max_tokens=128000,
     embed_dim=128,
@@ -477,6 +462,7 @@ evoqwen25_vl_retriever_7b_v1 = ModelMeta(
     release_date="2025-11-04",
     modalities=["image", "text"],
     n_parameters=7_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=14400,
     max_tokens=128000,
     embed_dim=128,

mteb/models/model_implementations/colsmol_models.py

@@ -56,10 +56,11 @@ colsmol_256m = ModelMeta(
     name="vidore/colSmol-256M",
     model_type=["late-interaction"],
     languages=["eng-Latn"],
-    revision="530094e83a40ca4edcb5c9e5ddfa61a4b5ea0d2f",
+    revision="a59110fdf114638b8018e6c9a018907e12f14855",
     release_date="2025-01-22",
     modalities=["image", "text"],
     n_parameters=256_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=800,
     max_tokens=8192,
     embed_dim=128,
@@ -87,6 +88,7 @@ colsmol_500m = ModelMeta(
     release_date="2025-01-22",
     modalities=["image", "text"],
     n_parameters=500_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1200,
     max_tokens=8192,
     embed_dim=128,

mteb/models/model_implementations/conan_models.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import hashlib
 import json
 import logging
@@ -5,20 +7,24 @@ import os
 import random
 import string
 import time
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import numpy as np
 import requests
-from torch.utils.data import DataLoader
 
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta
-from mteb.types import Array, BatchedInput, PromptType
 
 from .bge_models import bge_full_data
 from .e5_instruct import E5_MISTRAL_TRAINING_DATA
 
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
+
+
 conan_zh_datasets = {
     "BQ",
     "LCQMC",
@@ -205,6 +211,7 @@ Conan_embedding_v2 = ModelMeta(
     embed_dim=3584,
     open_weights=False,
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     license="apache-2.0",
     reference="https://huggingface.co/TencentBAC/Conan-embedding-v2",

mteb/models/model_implementations/dino_models.py

@@ -1,13 +1,18 @@
-from typing import Any, Literal
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any, Literal
 
 import torch
-from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
 
 
 class DINOModel(AbsEncoder):
@@ -112,6 +117,7 @@ dinov2_small = ModelMeta(
     release_date="2023-07-18",
     modalities=["image"],
     n_parameters=22_100_000,
+    n_embedding_parameters=None,
     memory_usage_mb=84,
     max_tokens=None,
     embed_dim=384,
@@ -143,6 +149,7 @@ dinov2_base = ModelMeta(
     release_date="2023-07-18",
     modalities=["image"],
     n_parameters=86_600_000,
+    n_embedding_parameters=None,
     memory_usage_mb=330,
     max_tokens=None,
     embed_dim=768,
@@ -174,6 +181,7 @@ dinov2_large = ModelMeta(
     release_date="2023-07-18",
     modalities=["image"],
     n_parameters=304_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1161,
     max_tokens=None,
     embed_dim=1024,
@@ -205,6 +213,7 @@ dinov2_giant = ModelMeta(
     release_date="2023-07-18",
     modalities=["image"],
     n_parameters=1_140_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=4335,
     max_tokens=None,
     embed_dim=1536,
@@ -240,6 +249,7 @@ webssl_dino300m_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=304_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1158,
     max_tokens=None,
     embed_dim=1024,
@@ -271,6 +281,7 @@ webssl_dino1b_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=1_130_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=4329,
     max_tokens=None,
     embed_dim=1536,
@@ -302,6 +313,7 @@ webssl_dino2b_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=2_080_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=7951,
     max_tokens=None,
     embed_dim=2688,
@@ -333,6 +345,7 @@ webssl_dino3b_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=3_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=11247,
     max_tokens=None,
     embed_dim=3072,
@@ -364,6 +377,7 @@ webssl_dino5b_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=5_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=18838,
     max_tokens=None,
     embed_dim=3584,
@@ -395,6 +409,7 @@ webssl_dino7b_full8b_224 = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=7_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=24605,
     max_tokens=None,
     embed_dim=4096,
@@ -426,6 +441,7 @@ webssl_dino7b_full8b_378 = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=7_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=24613,
     max_tokens=None,
     embed_dim=4096,
@@ -457,6 +473,7 @@ webssl_dino7b_full8b_518 = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=7_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=24623,
     max_tokens=None,
     embed_dim=4096,
@@ -489,6 +506,7 @@ webssl_dino2b_light2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=2_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=7951,
     max_tokens=None,
     embed_dim=2688,
@@ -520,6 +538,7 @@ webssl_dino2b_heavy2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=2_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=7951,
     max_tokens=None,
     embed_dim=2688,
@@ -551,6 +570,7 @@ webssl_dino3b_light2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=3_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=11247,
     max_tokens=None,
     embed_dim=3072,
@@ -582,6 +602,7 @@ webssl_dino3b_heavy2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=3_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=11247,
     max_tokens=None,
     embed_dim=3072,
@@ -613,6 +634,7 @@ webssl_mae300m_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=304_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1161,
     max_tokens=None,
     embed_dim=1024,
@@ -644,6 +666,7 @@ webssl_mae700m_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=700_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=2412,
     max_tokens=None,
     embed_dim=1280,
@@ -675,6 +698,7 @@ webssl_mae1b_full2b = ModelMeta(
     release_date="2025-04-24",
     modalities=["image"],
     n_parameters=1_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=4337,
     max_tokens=None,
     embed_dim=1536,

mteb/models/model_implementations/e5_instruct.py

@@ -57,6 +57,7 @@ e5_instruct = ModelMeta(
     use_instructions=True,
     reference="https://huggingface.co/intfloat/multilingual-e5-large-instruct",
     n_parameters=560_000_000,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=1068,
     embed_dim=1024,
     license="mit",
@@ -102,6 +103,7 @@ e5_mistral = ModelMeta(
     use_instructions=True,
     reference="https://huggingface.co/intfloat/e5-mistral-7b-instruct",
     n_parameters=7_111_000_000,
+    n_embedding_parameters=131_072_000,
     memory_usage_mb=13563,
     embed_dim=4096,
     license="mit",
@@ -145,6 +147,7 @@ zeta_alpha_ai__zeta_alpha_e5_mistral = ModelMeta(
     release_date="2024-08-30",
     languages=["eng-Latn"],
     n_parameters=7110660096,
+    n_embedding_parameters=None,
     memory_usage_mb=13563,
     max_tokens=32768.0,
     embed_dim=4096,
@@ -228,6 +231,7 @@ BeastyZ__e5_R_mistral_7b = ModelMeta(
     release_date="2024-06-28",
     languages=["eng-Latn"],
     n_parameters=7241732096,
+    n_embedding_parameters=131_072_000,
     memory_usage_mb=27625,
     max_tokens=32768.0,
     embed_dim=4096,

mteb/models/model_implementations/e5_models.py

@@ -76,6 +76,7 @@ e5_mult_small = ModelMeta(
     revision="fd1525a9fd15316a2d503bf26ab031a61d056e98",
     release_date=E5_PAPER_RELEASE_DATE,
     n_parameters=118_000_000,
+    n_embedding_parameters=96_014_208,
     memory_usage_mb=449,
     embed_dim=384,
     license="mit",
@@ -103,6 +104,7 @@ e5_mult_base = ModelMeta(
     revision="d13f1b27baf31030b7fd040960d60d909913633f",
     release_date=E5_PAPER_RELEASE_DATE,
     n_parameters=278_000_000,
+    n_embedding_parameters=192_001_536,
     memory_usage_mb=1061,
     embed_dim=768,
     license="mit",
@@ -130,6 +132,7 @@ e5_mult_large = ModelMeta(
     revision="ab10c1a7f42e74530fe7ae5be82e6d4f11a719eb",
     release_date=E5_PAPER_RELEASE_DATE,
     n_parameters=560_000_000,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=2136,
     embed_dim=1024,
     license="mit",
@@ -157,6 +160,7 @@ e5_eng_small_v2 = ModelMeta(
     revision="dca8b1a9dae0d4575df2bf423a5edb485a431236",
     release_date=E5_PAPER_RELEASE_DATE,
     n_parameters=33_000_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=127,
     embed_dim=384,
     license="mit",
@@ -184,6 +188,7 @@ e5_eng_small = ModelMeta(
     revision="e272f3049e853b47cb5ca3952268c6662abda68f",
     release_date=E5_PAPER_RELEASE_DATE,
     n_parameters=33_000_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=127,
     embed_dim=384,
     license="mit",
@@ -211,6 +216,7 @@ e5_eng_base_v2 = ModelMeta(
     revision="1c644c92ad3ba1efdad3f1451a637716616a20e8",
     release_date=E5_PAPER_RELEASE_DATE,
     n_parameters=109_000_000,
+    n_embedding_parameters=23_440_896,
     memory_usage_mb=418,
     embed_dim=768,
     license="mit",
@@ -239,6 +245,7 @@ e5_eng_large_v2 = ModelMeta(
     revision="b322e09026e4ea05f42beadf4d661fb4e101d311",
     release_date=E5_PAPER_RELEASE_DATE,
     n_parameters=335_000_000,
+    n_embedding_parameters=31_254_528,
     memory_usage_mb=1278,
     embed_dim=1024,
     license="mit",
@@ -267,6 +274,7 @@ e5_large = ModelMeta(
     revision="4dc6d853a804b9c8886ede6dda8a073b7dc08a81",
     release_date="2022-12-26",
     n_parameters=335_000_000,
+    n_embedding_parameters=31_254_528,
     memory_usage_mb=1278,
     embed_dim=1024,
     license="apache-2.0",
@@ -295,6 +303,7 @@ e5_base = ModelMeta(
     revision="b533fe4636f4a2507c08ddab40644d20b0006d6a",
     release_date="2022-12-26",
     n_parameters=109_000_000,
+    n_embedding_parameters=23_440_896,
     memory_usage_mb=418,
     embed_dim=768,
     license="apache-2.0",

mteb/models/model_implementations/e5_v.py

@@ -1,14 +1,19 @@
-from typing import Any
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
 
 import torch
 from packaging import version
-from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
 
 E5_V_TRANSFORMERS_VERSION = (
     "4.44.2"  # Issue 1647: Only works with transformers==4.44.2.
@@ -166,6 +171,7 @@ e5_v = ModelMeta(
     release_date="2024-07-17",
     modalities=["image", "text"],
     n_parameters=8_360_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=15936,
     max_tokens=8192,
     embed_dim=4096,

mteb/models/model_implementations/eagerworks_models.py

@@ -1,17 +1,23 @@
-from typing import Any
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
 
 import torch
-from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
 from mteb._requires_package import (
     requires_image_dependencies,
     requires_package,
 )
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
+from mteb.types import PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput
 
 
 class EagerEmbedV1Wrapper(AbsEncoder):
@@ -147,6 +153,7 @@ Eager_Embed_V1 = ModelMeta(
     release_date="2025-11-20",
     modalities=["image", "text"],
     n_parameters=4_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=16929,
     max_tokens=262144,
     embed_dim=2560,

mteb/models/model_implementations/emillykkejensen_models.py

@@ -10,6 +10,7 @@ embedding_gemma_300m_scandi = ModelMeta(
     revision="9f3307b9f601db564a9190cb475324d128dcfe86",
     release_date="2025-10-17",
     n_parameters=307_581_696,
+    n_embedding_parameters=None,
     embed_dim=768,
     max_tokens=2048,
     license="apache-2.0",
@@ -43,6 +44,7 @@ qwen_scandi = ModelMeta(
     revision="cf1e7ba36ebd3d605549d8f02930a18e17b54513",
     release_date="2025-10-17",
     n_parameters=595776512,
+    n_embedding_parameters=None,
     memory_usage_mb=2272,
     embed_dim=1024,
     max_tokens=32768,
@@ -67,6 +69,7 @@ mmbert_scandi = ModelMeta(
     revision="82d74c7a5d8e1ddf31b132865df2d16b2b0294ee",
     release_date="2025-10-17",
     n_parameters=306939648,
+    n_embedding_parameters=None,
     memory_usage_mb=1171,
     embed_dim=768,
     max_tokens=8192,

mteb/models/model_implementations/en_code_retriever.py

@@ -18,6 +18,7 @@ english_code_retriever = ModelMeta(
     revision="be653fab7d27a7348a0c2c3d16b9f92a7f10cb0c",
     release_date="2025-07-10",
     n_parameters=149_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=568,
     embed_dim=768,
     license="mit",

mteb/models/model_implementations/euler_models.py

@@ -9,6 +9,7 @@ Euler_Legal_Embedding_V1 = ModelMeta(
     release_date="2025-11-06",
     languages=["eng-Latn"],
     n_parameters=8000000000,
+    n_embedding_parameters=None,
     memory_usage_mb=15618,
     max_tokens=1536,
     embed_dim=4096,

mteb/models/model_implementations/evaclip_models.py

@@ -1,15 +1,20 @@
+from __future__ import annotations
+
 from pathlib import Path
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import torch
-from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
 from mteb._requires_package import requires_image_dependencies
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
 
 EVA_CLIP_CITATION = """@article{EVA-CLIP,
       title={EVA-CLIP: Improved Training Techniques for CLIP at Scale},
@@ -144,6 +149,7 @@ EVA02_CLIP_B_16 = ModelMeta(
     release_date="2023-04-26",
     modalities=["image", "text"],
     n_parameters=149_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=568,
     max_tokens=77,
     embed_dim=512,
@@ -168,6 +174,7 @@ EVA02_CLIP_L_14 = ModelMeta(
     release_date="2023-04-26",
     modalities=["image", "text"],
     n_parameters=428_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1633,
     max_tokens=77,
     embed_dim=768,
@@ -192,6 +199,7 @@ EVA02_CLIP_bigE_14 = ModelMeta(
     release_date="2023-04-26",
     modalities=["image", "text"],
     n_parameters=4_700_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=17929,
     max_tokens=77,
     embed_dim=1024,
@@ -217,6 +225,7 @@ EVA02_CLIP_bigE_14_plus = ModelMeta(
     release_date="2023-04-26",
     modalities=["image", "text"],
     n_parameters=5_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=19073,
     max_tokens=77,
     embed_dim=1024,

mteb/models/model_implementations/fa_models.py

@@ -12,6 +12,7 @@ parsbert = ModelMeta(
     revision="d73a0e2c7492c33bd5819bcdb23eba207404dd19",
     release_date="2021-05-19",
     n_parameters=162_841_344,
+    n_embedding_parameters=76_800_000,
     memory_usage_mb=621,
     embed_dim=768,
     license="not specified",
@@ -48,6 +49,7 @@ bert_zwnj = ModelMeta(
     revision="b9506ddc579ac8c398ae6dae680401ae0a1a5b23",
     release_date="2021-06-28",
     n_parameters=118_297_344,
+    n_embedding_parameters=32_256_000,
     memory_usage_mb=451,
     embed_dim=768,
     license="not specified",
@@ -74,6 +76,7 @@ roberta_zwnj = ModelMeta(
     revision="36f912ac44e22250aee16ea533a4ff8cd848c1a1",
     release_date="2021-06-28",
     n_parameters=118_298_112,
+    n_embedding_parameters=32_256_000,
     memory_usage_mb=451,
     embed_dim=768,
     license="not specified",
@@ -99,6 +102,7 @@ sentence_transformer_parsbert = ModelMeta(
     revision="72bd0a3557622f0ae08a092f4643609e0b950cdd",
     release_date="2024-12-10",
     n_parameters=162_841_344,
+    n_embedding_parameters=76_800_000,
     memory_usage_mb=621,
     embed_dim=768,
     license="apache-2.0",
@@ -123,6 +127,7 @@ tooka_bert_base = ModelMeta(
     revision="fa5ca89df5670700d9325b8872ac65c17cb24582",
     release_date="2024-12-08",
     n_parameters=122_905_344,
+    n_embedding_parameters=36_864_000,
     memory_usage_mb=469,
     embed_dim=768,
     license="apache-2.0",
@@ -150,6 +155,7 @@ tooka_sbert = ModelMeta(
     revision="5d07f0c543aca654373b931ae07cd197769110fd",
     release_date="2024-12-07",
     n_parameters=353_039_360,
+    n_embedding_parameters=49_152_000,
     memory_usage_mb=1347,
     embed_dim=1024,
     license="apache-2.0",
@@ -181,6 +187,7 @@ fa_bert = ModelMeta(
     revision="a0e3973064c97768e121b9b95f21adc94e0ca3fb",
     release_date="2024-10-07",
     n_parameters=124_441_344,
+    n_embedding_parameters=38_400_000,
     memory_usage_mb=475,
     embed_dim=768,
     license="not specified",
@@ -229,6 +236,7 @@ tooka_sbert_v2_small = ModelMeta(
     revision="8bbed87e36669387f71437c061430ba56d1b496f",
     release_date="2025-05-01",
     n_parameters=122_905_344,
+    n_embedding_parameters=36_864_000,
     memory_usage_mb=496,
     embed_dim=768,
     license="not specified",
@@ -260,6 +268,7 @@ tooka_sbert_v2_large = ModelMeta(
     revision="b59682efa961122cc0e4408296d5852870c82eae",
     release_date="2025-05-01",
     n_parameters=353_039_360,
+    n_embedding_parameters=49_152_000,
     memory_usage_mb=1347,
     embed_dim=1024,
     license="not specified",