mteb 2.5.2__py3-none-any.whl → 2.7.9__py3-none-any.whl

mteb/models/model_implementations/e5_models.py

@@ -76,13 +76,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/intfloat/multilingual-e5-small",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     use_instructions=True,
     public_training_code=None,  # couldn't find
     public_training_data=None,
@@ -103,13 +104,14 @@ 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",
     max_tokens=514,
     reference="https://huggingface.co/intfloat/multilingual-e5-base",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data=None,
@@ -130,13 +132,14 @@ 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",
     max_tokens=514,
     reference="https://huggingface.co/intfloat/multilingual-e5-large",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data=None,
@@ -157,13 +160,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/intfloat/e5-small-v2",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data=None,
@@ -184,13 +188,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/intfloat/e5-small",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data=None,
@@ -211,13 +216,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/intfloat/e5-base-v2",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     use_instructions=True,
     superseded_by=None,
     adapted_from="intfloat/e5-base",
@@ -239,13 +245,14 @@ 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",
     max_tokens=514,
     reference="https://huggingface.co/intfloat/e5-large-v2",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     use_instructions=True,
     superseded_by=None,
     adapted_from="intfloat/e5-large",
@@ -267,13 +274,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/intfloat/e5-large",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=True,
     superseded_by="intfloat/e5-large-v2",
     adapted_from="google-bert/bert-large-uncased-whole-word-masking",
@@ -295,13 +303,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/intfloat/e5-base",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=True,
     superseded_by="intfloat/e5-base-v2",
     adapted_from="google-bert/bert-base-uncased",

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.
@@ -30,6 +35,7 @@ class E5VModel(AbsEncoder):
         self,
         model_name: str,
         revision: str,
+        device: str | None = None,
         composed_prompt=None,
         **kwargs: Any,
     ):
@@ -47,8 +53,7 @@ class E5VModel(AbsEncoder):
         self.processor = LlavaNextProcessor.from_pretrained(
             model_name, revision=revision
         )
-        if "device" in kwargs:
-            self.device = kwargs.pop("device")
+        self.device = device or ("cuda" if torch.cuda.is_available() else "cpu")
         self.model = LlavaNextForConditionalGeneration.from_pretrained(
             model_name, revision=revision, **kwargs
         )
@@ -87,7 +92,7 @@ class E5VModel(AbsEncoder):
                     ],
                     return_tensors="pt",
                     padding=True,
-                ).to("cuda")
+                ).to(self.device)
                 text_outputs = self.model(
                     **text_inputs, output_hidden_states=True, return_dict=True
                 ).hidden_states[-1][:, -1, :]
@@ -111,7 +116,7 @@ class E5VModel(AbsEncoder):
                     batch["image"],
                     return_tensors="pt",
                     padding=True,
-                ).to("cuda")
+                ).to(self.device)
                 image_outputs = self.model(
                     **img_inputs, output_hidden_states=True, return_dict=True
                 ).hidden_states[-1][:, -1, :]
@@ -141,7 +146,7 @@ class E5VModel(AbsEncoder):
                     ]
                     inputs = self.processor(
                         prompts, batch["image"], return_tensors="pt", padding=True
-                    ).to("cuda")
+                    ).to(self.device)
                     outputs = self.model(
                         **inputs, output_hidden_states=True, return_dict=True
                     ).hidden_states[-1][:, -1, :]
@@ -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,
@@ -173,7 +179,7 @@ e5_v = ModelMeta(
     open_weights=True,
     public_training_code="https://github.com/kongds/E5-V",
     public_training_data="https://huggingface.co/datasets/princeton-nlp/datasets-for-simcse",
-    framework=["PyTorch"],
+    framework=["PyTorch", "Transformers", "safetensors"],
     reference="https://huggingface.co/royokong/e5-v",
     similarity_fn_name=ScoringFunction.COSINE,
     use_instructions=True,

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):
@@ -147,12 +153,13 @@ 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,
     license="apache-2.0",
     open_weights=True,
-    framework=["Tevatron"],
+    framework=["Tevatron", "safetensors"],
     reference="https://huggingface.co/eagerworks/eager-embed-v1",
     similarity_fn_name=ScoringFunction.COSINE,
     use_instructions=True,

mteb/models/model_implementations/emillykkejensen_models.py

@@ -2,7 +2,7 @@ from mteb.models.model_meta import ModelMeta
 from mteb.models.sentence_transformer_wrapper import sentence_transformers_loader
 
 embedding_gemma_300m_scandi = ModelMeta(
-    loader=sentence_transformers_loader,  # type: ignore
+    loader=sentence_transformers_loader,
     name="emillykkejensen/EmbeddingGemma-Scandi-300m",
     model_type=["dense"],
     languages=["dan-Latn", "swe-Latn", "nor-Latn", "nob-Latn", "nno-Latn"],
@@ -10,11 +10,12 @@ 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",
     reference="https://huggingface.co/emillykkejensen/EmbeddingGemma-Scandi-300m",
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data="https://huggingface.co/datasets/DDSC/nordic-embedding-training-data",
@@ -35,7 +36,7 @@ embedding_gemma_300m_scandi = ModelMeta(
 
 
 qwen_scandi = ModelMeta(
-    loader=sentence_transformers_loader,  # type: ignore
+    loader=sentence_transformers_loader,
     name="emillykkejensen/Qwen3-Embedding-Scandi-0.6B",
     model_type=["dense"],
     languages=["dan-Latn", "swe-Latn", "nor-Latn", "nob-Latn", "nno-Latn"],
@@ -43,12 +44,13 @@ 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,
     license="apache-2.0",
     reference="https://huggingface.co/emillykkejensen/Qwen3-Embedding-Scandi-0.6B",
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data="https://huggingface.co/datasets/DDSC/nordic-embedding-training-data",
@@ -59,7 +61,7 @@ qwen_scandi = ModelMeta(
 
 
 mmbert_scandi = ModelMeta(
-    loader=sentence_transformers_loader,  # type: ignore
+    loader=sentence_transformers_loader,
     name="emillykkejensen/mmBERTscandi-base-embedding",
     model_type=["dense"],
     languages=["dan-Latn", "swe-Latn", "nor-Latn", "nob-Latn", "nno-Latn"],
@@ -67,12 +69,13 @@ 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,
     license="apache-2.0",
     reference="https://huggingface.co/emillykkejensen/Qwen3-Embedding-Scandi-0.6B",
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data="https://huggingface.co/datasets/DDSC/nordic-embedding-training-data",

mteb/models/model_implementations/en_code_retriever.py

@@ -18,13 +18,14 @@ 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",
     max_tokens=8192,
     reference="https://huggingface.co/fyaronskiy/english_code_retriever",
     similarity_fn_name="cosine",
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=True,
     public_training_code=None,
     public_training_data="https://huggingface.co/datasets/code-search-net/code_search_net",

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,
@@ -16,7 +17,7 @@ Euler_Legal_Embedding_V1 = ModelMeta(
     open_weights=True,
     public_training_code=None,
     public_training_data=None,
-    framework=["PyTorch", "Sentence Transformers"],
+    framework=["PyTorch", "Sentence Transformers", "safetensors", "Transformers"],
     reference="https://huggingface.co/Mira190/Euler-Legal-Embedding-V1",
     similarity_fn_name="cosine",
     use_instructions=False,
@@ -24,7 +25,7 @@ Euler_Legal_Embedding_V1 = ModelMeta(
     adapted_from="Qwen/Qwen3-Embedding-8B",
     superseded_by=None,
     citation="""@misc{euler2025legal,
-      title={Euler-Legal-Embedding: Advanced Legal Representation Learning}, 
+      title={Euler-Legal-Embedding: Advanced Legal Representation Learning},
       author={LawRank Team},
       year={2025},
       publisher={Hugging Face}

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

@@ -12,13 +12,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/HooshvareLab/bert-base-parsbert-uncased",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "Transformers"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -48,13 +49,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/m3hrdadfi/bert-zwnj-wnli-mean-tokens",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "Transformers"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -74,13 +76,14 @@ 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",
     max_tokens=514,
     reference="https://huggingface.co/m3hrdadfi/roberta-zwnj-wnli-mean-tokens",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "Transformers"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -99,13 +102,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/myrkur/sentence-transformer-parsbert-fa",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -123,13 +127,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/PartAI/TookaBERT-Base",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "Transformers", "safetensors"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -150,13 +155,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/PartAI/Tooka-SBERT",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -181,13 +187,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/sbunlp/fabert",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "Transformers", "safetensors"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -229,13 +236,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/PartAI/Tooka-SBERT-V2-Small",
     similarity_fn_name="cosine",
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -260,13 +268,14 @@ 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",
     max_tokens=512,
     reference="https://huggingface.co/PartAI/Tooka-SBERT-V2-Large",
     similarity_fn_name="cosine",
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors"],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,

mteb/models/model_implementations/facebookai.py

@@ -113,13 +113,20 @@ xlmr_base = ModelMeta(
     revision="e73636d4f797dec63c3081bb6ed5c7b0bb3f2089",
     release_date="2019-11-05",  # arxiv paper release
     n_parameters=278043648,
+    n_embedding_parameters=192_001_536,
     memory_usage_mb=1064,
     embed_dim=768,
     license="mit",
     max_tokens=512,
     reference="https://huggingface.co/FacebookAI/xlm-roberta-base",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=[
+        "Sentence Transformers",
+        "PyTorch",
+        "Transformers",
+        "ONNX",
+        "safetensors",
+    ],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,
@@ -157,13 +164,20 @@ xlmr_large = ModelMeta(
     revision="c23d21b0620b635a76227c604d44e43a9f0ee389",
     release_date="2019-11-05",  # arxiv paper release
     n_parameters=559890432,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=2141,
     embed_dim=1024,
     license="mit",
     max_tokens=512,
     reference="https://huggingface.co/FacebookAI/xlm-roberta-large",
     similarity_fn_name=ScoringFunction.COSINE,
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=[
+        "Sentence Transformers",
+        "PyTorch",
+        "Transformers",
+        "ONNX",
+        "safetensors",
+    ],
     use_instructions=False,
     public_training_code=None,
     public_training_data=None,

mteb/models/model_implementations/geogpt_models.py

@@ -20,13 +20,14 @@ geoembedding = ModelMeta(
     ),
     release_date="2025-04-22",
     n_parameters=7241732096,
+    n_embedding_parameters=131_072_000,
     memory_usage_mb=27625,
     embed_dim=4096,
     license="apache-2.0",
     max_tokens=32768,
     reference="https://huggingface.co/GeoGPT-Research-Project/GeoEmbedding",
     similarity_fn_name="cosine",
-    framework=["Sentence Transformers", "PyTorch"],
+    framework=["Sentence Transformers", "PyTorch", "safetensors", "Transformers"],
     use_instructions=True,
     public_training_code=None,
     public_training_data=None,