mteb 2.5.1__py3-none-any.whl → 2.5.3__py3-none-any.whl

mteb/abstasks/abstask.py

@@ -1,5 +1,6 @@
 import json
 import logging
+import warnings
 from abc import ABC, abstractmethod
 from collections.abc import Sequence
 from copy import copy
@@ -102,9 +103,9 @@ class AbsTask(ABC):
     def check_if_dataset_is_superseded(self) -> None:
         """Check if the dataset is superseded by a newer version."""
         if self.superseded_by:
-            logger.warning(
-                f"Dataset '{self.metadata.name}' is superseded by '{self.superseded_by}', you might consider using the newer version of the dataset."
-            )
+            msg = f"Dataset '{self.metadata.name}' is superseded by '{self.superseded_by}'. We recommend using the newer version of the dataset unless you are running a specific benchmark. See `get_task('{self.superseded_by}').metadata.description` to get a description of the task and changes."
+            logger.warning(msg)
+            warnings.warn(msg)
 
     def dataset_transform(self):
         """A transform operations applied to the dataset after loading.
@@ -607,9 +608,8 @@ class AbsTask(ABC):
             self.data_loaded = False
             logger.info(f"Unloaded dataset {self.metadata.name} from memory.")
         else:
-            logger.warning(
-                f"Dataset {self.metadata.name} is not loaded, cannot unload it."
-            )
+            msg = f"Dataset `{self.metadata.name}` is not loaded, cannot unload it."
+            logger.warning(msg)
 
     @property
     def superseded_by(self) -> str | None:

mteb/abstasks/aggregated_task.py

@@ -1,4 +1,5 @@
 import logging
+import warnings
 from pathlib import Path
 from typing import Any
 
@@ -113,20 +114,13 @@ class AbsTaskAggregate(AbsTask):
         )
         mteb_versions = {tr.mteb_version for tr in task_results}
         if len(mteb_versions) != 1:
-            logger.warning(
-                f"All tasks of {self.metadata.name} is not run using the same version."
-            )
+            msg = f"All tasks of {self.metadata.name} is not run using the same version. different versions found are: {mteb_versions}"
+            logger.warning(msg)
+            warnings.warn(msg)
             task_res.mteb_version = None
         task_res.mteb_version = task_results[0].mteb_version
         return task_res
 
-    def check_if_dataset_is_superseded(self) -> None:
-        """Check if the dataset is superseded by a newer version"""
-        if self.superseded_by:
-            logger.warning(
-                f"Dataset '{self.metadata.name}' is superseded by '{self.superseded_by}', you might consider using the newer version of the dataset."
-            )
-
     def filter_eval_splits(self, eval_splits: list[str] | None) -> Self:
         """Filter the evaluation splits of the task.
 

mteb/abstasks/clustering_legacy.py

@@ -89,6 +89,9 @@ class AbsTaskClusteringLegacy(AbsTask):
         prediction_folder: Path | None = None,
         **kwargs: Any,
     ) -> ScoresDict:
+        data_split = data_split.select_columns(
+            [self.input_column_name, self.label_column_name]
+        )
         # MTEB text clustering requires renaming and eval per subset.
         if self.metadata.modalities == ["text"]:
             all_metrics = []
@@ -97,8 +100,6 @@ class AbsTaskClusteringLegacy(AbsTask):
                 logger.info(
                     f"Running clustering on cluster ({i + 1}/{len(data_split)})"
                 )
-                if "__index_level_0__" in cluster_set:
-                    cluster_set.pop("__index_level_0__")
                 clustering_dataset = Dataset.from_dict(cluster_set).select_columns(
                     [self.input_column_name, self.label_column_name]
                 )

mteb/abstasks/task_metadata.py

@@ -376,9 +376,8 @@ class TaskMetadata(BaseModel):
         if include_cite and cite:
             # check for whitespace in the citation
             if " " in cite:
-                logger.warning(
-                    "Citation contains whitespace. Please ensure that the citation is correctly formatted."
-                )
+                msg = "Citation contains whitespace. Please ensure that the citation is correctly formatted."
+                logger.warning(msg)
             return f"\\cite{{{cite}}}"
         return cite
 

mteb/cache.py

@@ -3,6 +3,7 @@ import logging
 import os
 import shutil
 import subprocess
+import warnings
 from collections import defaultdict
 from collections.abc import Sequence
 from pathlib import Path
@@ -83,9 +84,9 @@ class ResultCache:
         model_path = results_folder / model_name
 
         if model_revision is None:
-            logger.warning(
-                "model_revision is not specified, attempting to load the latest revision. To disable this behavior, specify model_revision explicitly."
-            )
+            msg = "`model_revision` is not specified, attempting to load the latest revision. To disable this behavior, specify the 'model_revision` explicitly."
+            logger.warning(msg)
+            warnings.warn(msg)
             # get revs from paths
             revisions = [p for p in model_path.glob("*") if p.is_dir()]
             if not revisions:
@@ -281,7 +282,9 @@ class ResultCache:
             shutil.rmtree(self.cache_path)
             logger.info(f"Cache directory {self.cache_path} cleared.")
         else:
-            logger.warning(f"Cache directory {self.cache_path} does not exist.")
+            msg = f"Cache directory `{self.cache_path}` does not exist."
+            logger.warning(msg)
+            warnings.warn(msg)
 
     def __repr__(self) -> str:
         return f"ResultCache(cache_path={self.cache_path})"

mteb/cli/build_cli.py

@@ -1,6 +1,7 @@
 import argparse
 import logging
 import os
+import warnings
 from pathlib import Path
 
 import torch
@@ -69,15 +70,17 @@ def run(args: argparse.Namespace) -> None:
 
     overwrite_strategy = args.overwrite_strategy
     if args.overwrite:
-        logger.warning(
-            "`--overwrite` is deprecated, please use `--overwrite-strategy 'always'` instead."
+        warnings.warn(
+            "`--overwrite` is deprecated, please use `--overwrite-strategy 'always'` instead.",
+            DeprecationWarning,
         )
         overwrite_strategy = OverwriteStrategy.ALWAYS.value
 
     prediction_folder = args.prediction_folder
     if args.save_predictions:
-        logger.warning(
-            "`--save_predictions` is deprecated, please use `--prediction-folder` instead."
+        warnings.warn(
+            "`--save_predictions` is deprecated, please use `--prediction-folder` instead.",
+            DeprecationWarning,
         )
         prediction_folder = args.output_folder
 
@@ -279,7 +282,9 @@ def _create_meta(args: argparse.Namespace) -> None:
         from_existing = Path(from_existing)
 
     if output_path.exists() and overwrite:
-        logger.warning("Output path already exists, overwriting.")
+        msg = "Output path already exists, overwriting."
+        logger.warning(msg)
+        warnings.warn(msg)
     elif output_path.exists():
         raise FileExistsError(
             "Output path already exists, use --overwrite to overwrite."

mteb/cli/generate_model_card.py

@@ -1,4 +1,5 @@
 import logging
+import warnings
 from pathlib import Path
 
 from huggingface_hub import ModelCard, ModelCardData, repo_exists
@@ -92,9 +93,9 @@ def generate_model_card(
         if repo_exists(existing_model_card_id_or_path):
             existing_model_card.push_to_hub(existing_model_card_id_or_path, token=token)
         else:
-            logger.warning(
-                f"Repository {existing_model_card_id_or_path} does not exist on the Hub. Skipping push to hub."
-            )
+            msg = f"Repository {existing_model_card_id_or_path} does not exist on the Hub. Skipping push to hub."
+            logger.warning(msg)
+            warnings.warn(msg)
     existing_model_card.save(output_path)
 
 

mteb/deprecated_evaluator.py

@@ -5,6 +5,7 @@ import logging
 import os
 import sys
 import traceback
+import warnings
 from collections.abc import Iterable
 from copy import deepcopy
 from datetime import datetime
@@ -470,9 +471,9 @@ class MTEB:
                             raise ImportError(
                                 "codecarbon is not installed. Please install it using `pip install 'mteb[codecarbon]'` to track CO₂ emissions."
                             )
-                        logger.warning(
-                            "Evaluating multiple MTEB runs simultaneously will produce incorrect CO₂ results"
-                        )
+                        msg = "Evaluating multiple MTEB runs simultaneously will produce incorrect CO₂ results"
+                        logger.warning(msg)
+                        warnings.warn(msg)
                         with EmissionsTracker(
                             save_to_file=False,
                             save_to_api=False,

mteb/evaluate.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import logging
+import warnings
 from collections.abc import Iterable
 from pathlib import Path
 from time import time
@@ -136,10 +137,12 @@ def _evaluate_task(
             task.load_data()
         except DatasetNotFoundError as e:
             if not task.metadata.is_public and public_only is None:
-                logger.warning(
+                msg = (
                     f"Dataset for private task '{task.metadata.name}' not found. "
                     "Make sure you have access to the dataset and that you have set up the authentication correctly. To disable this warning set `public_only=False`"
                 )
+                logger.warning(msg)
+                warnings.warn(msg)
                 return TaskError(
                     task_name=task.metadata.name,
                     exception=str(e),

mteb/get_tasks.py

@@ -2,6 +2,7 @@
 
 import difflib
 import logging
+import warnings
 from collections import Counter, defaultdict
 from collections.abc import Sequence
 from typing import Any
@@ -340,9 +341,9 @@ def get_task(
     """
     if task_name in _TASK_RENAMES:
         _task_name = _TASK_RENAMES[task_name]
-        logger.warning(
-            f"The task with the given name '{task_name}' has been renamed to '{_task_name}'. To prevent this warning use the new name."
-        )
+        msg = f"The task with the given name '{task_name}' has been renamed to '{_task_name}'. To prevent this warning use the new name."
+        logger.warning(msg)
+        warnings.warn(msg)
 
     if task_name not in _TASKS_REGISTRY:
         close_matches = difflib.get_close_matches(task_name, _TASKS_REGISTRY.keys())

mteb/leaderboard/app.py

@@ -36,9 +36,15 @@ LANGUAGE: list[str] = list({l for t in mteb.get_tasks() for l in t.metadata.lang
 
 
 def _load_results(cache: ResultCache) -> BenchmarkResults:
+    start_time = time.time()
     results_cache_path = Path(__file__).parent.joinpath("__cached_results.json")
     if not results_cache_path.exists():
+        logger.info("Cached results not found, downloading from remote...")
         cache.download_from_remote()
+        download_time = time.time() - start_time
+        logger.info(f"Downloaded remote results in {download_time:.2f}s")
+
+        load_start = time.time()
         all_model_names = [model_meta.name for model_meta in mteb.get_model_metas()]
 
         all_results = cache.load_results(
@@ -47,10 +53,16 @@ def _load_results(cache: ResultCache) -> BenchmarkResults:
             require_model_meta=False,
             include_remote=True,
         )
+        load_time = time.time() - load_start
+        logger.info(f"Loaded results from cache in {load_time:.2f}s")
         return all_results
     else:
+        logger.info("Loading cached results from disk...")
         with results_cache_path.open() as cache_file:
-            return mteb.BenchmarkResults.from_validated(**json.load(cache_file))
+            results = mteb.BenchmarkResults.from_validated(**json.load(cache_file))
+        total_time = time.time() - start_time
+        logger.info(f"Loaded cached results in {total_time:.2f}s")
+        return results
 
 
 def _produce_benchmark_link(benchmark_name: str, request: gr.Request) -> str:
@@ -322,20 +334,48 @@ def _cache_update_task_list(
 
 def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
     """Returns a Gradio Blocks app for the MTEB leaderboard."""
-    logger.info("Loading all benchmark results")
+    app_start = time.time()
+    logger.info("=== Starting leaderboard app initialization ===")
+
+    logger.info("Step 1/7: Loading all benchmark results...")
+    load_start = time.time()
     all_results = _load_results(cache)
+    load_time = time.time() - load_start
+    logger.info(f"Step 1/7 complete: Loaded results in {load_time:.2f}s")
 
+    logger.info("Step 2/7: Fetching benchmarks...")
+    bench_start = time.time()
     benchmarks = sorted(
         mteb.get_benchmarks(display_on_leaderboard=True), key=lambda x: x.name
     )
+    bench_time = time.time() - bench_start
+    logger.info(
+        f"Step 2/7 complete: Fetched {len(benchmarks)} benchmarks in {bench_time:.2f}s"
+    )
+
+    logger.info(
+        "Step 3/7: Processing all benchmarks (select_tasks + join_revisions)..."
+    )
+    process_start = time.time()
     all_benchmark_results = {
         benchmark.name: all_results.select_tasks(benchmark.tasks).join_revisions()
         for benchmark in benchmarks
     }
+    process_time = time.time() - process_start
+    if len(benchmarks) > 0:
+        logger.info(
+            f"Step 3/7 complete: Processed {len(benchmarks)} benchmarks in {process_time:.2f}s (avg {process_time / len(benchmarks):.2f}s/benchmark)"
+        )
+    else:
+        logger.info(
+            f"Step 3/7 complete: Processed 0 benchmarks in {process_time:.2f}s (avg N/A)"
+        )
 
     default_benchmark = mteb.get_benchmark(DEFAULT_BENCHMARK_NAME)
     default_results = all_benchmark_results[default_benchmark.name]
-    logger.info("Benchmark results loaded")
+
+    logger.info("Step 4/7: Filtering models...")
+    filter_start = time.time()
 
     default_scores = default_results._get_scores(format="long")
     all_models = list({entry["model_name"] for entry in default_scores})
@@ -355,7 +395,13 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
     # Filter BenchmarkResults based on default filtered models (as required by Kenneth)
     filtered_model_names = [entry["model_name"] for entry in default_filtered_scores]
     filtered_benchmark_results = default_results.select_models(filtered_model_names)
+    filter_time = time.time() - filter_start
+    logger.info(
+        f"Step 4/7 complete: Filtered {len(filtered_model_names)} models in {filter_time:.2f}s"
+    )
 
+    logger.info("Step 5/7: Generating tables...")
+    table_start = time.time()
     summary_table = apply_summary_styling_from_benchmark(
         default_benchmark, filtered_benchmark_results
     )
@@ -366,10 +412,14 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
         default_benchmark,
         filtered_benchmark_results,
     )
+    table_time = time.time() - table_start
+    logger.info(f"Step 5/7 complete: Generated tables in {table_time:.2f}s")
 
     # Check if this benchmark displays per-language results
     display_language_table = len(default_benchmark.language_view) > 0
 
+    logger.info("Step 6/7: Creating Gradio components...")
+    component_start = time.time()
     lang_select = gr.CheckboxGroup(
         sorted(default_results.languages),
         value=sorted(default_results.languages),
@@ -410,7 +460,13 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
         label="Modality",
         info="Select modalities to include.",
     )
+    component_time = time.time() - component_start
+    logger.info(
+        f"Step 6/7 complete: Created Gradio components in {component_time:.2f}s"
+    )
 
+    logger.info("Step 7/7: Building Gradio interface and callbacks...")
+    interface_start = time.time()
     with gr.Blocks(fill_width=True) as demo:
         with gr.Sidebar(
             position="left",
@@ -926,7 +982,11 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
             )
 
         gr.Markdown(ACKNOWLEDGEMENT, elem_id="ack_markdown")
+    interface_time = time.time() - interface_start
+    logger.info(f"Step 7/7 complete: Built Gradio interface in {interface_time:.2f}s")
 
+    logger.info("Starting prerun on all benchmarks to populate caches...")
+    prerun_start = time.time()
     # Prerun on all benchmarks, so that results of callbacks get cached
     for benchmark in benchmarks:
         (
@@ -952,6 +1012,13 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
         update_tables(
             bench_scores, filtered_tasks, bench_initial_models, benchmark.name
         )
+    prerun_time = time.time() - prerun_start
+    logger.info(
+        f"Prerun complete: Processed {len(benchmarks)} benchmarks in {prerun_time:.2f}s"
+    )
+
+    total_time = time.time() - app_start
+    logger.info(f"=== Leaderboard app initialization complete in {total_time:.2f}s ===")
     return demo
 
 

mteb/models/abs_encoder.py

@@ -1,4 +1,5 @@
 import logging
+import warnings
 from abc import ABC, abstractmethod
 from collections.abc import Callable, Sequence
 from typing import Any, Literal, cast, get_args, overload
@@ -187,6 +188,7 @@ class AbsEncoder(ABC):
                 except KeyError:
                     msg = f"Task name {task_name} is not valid. {valid_keys_msg}"
                     logger.warning(msg)
+                    warnings.warn(msg)
                     invalid_task_messages.add(msg)
                     invalid_keys.add(task_key)
 
@@ -232,9 +234,9 @@ class AbsEncoder(ABC):
         if isinstance(prompt, dict) and prompt_type:
             if prompt.get(prompt_type.value):
                 return prompt[prompt_type.value]
-            logger.warning(
-                f"Prompt type '{prompt_type}' not found in task metadata for task '{task_metadata.name}'."
-            )
+            msg = f"Prompt type '{prompt_type}' not found in task metadata for task '{task_metadata.name}'."
+            logger.warning(msg)
+            warnings.warn(msg)
             return ""
 
         if prompt:

mteb/models/cache_wrappers/cache_backends/faiss_cache.py

@@ -1,5 +1,6 @@
 import json
 import logging
+import warnings
 from pathlib import Path
 
 import numpy as np
@@ -71,7 +72,9 @@ class FaissCache:
         try:
             return self.index.reconstruct(idx)
         except Exception:
-            logger.warning(f"Vector id {idx} missing for hash {item_hash}")
+            msg = f"Vector id {idx} missing for hash {item_hash}"
+            logger.warning(msg)
+            warnings.warn(msg)
             return None
 
     def save(self) -> None:

mteb/models/cache_wrappers/cache_backends/numpy_cache.py

@@ -1,5 +1,6 @@
 import json
 import logging
+import warnings
 from pathlib import Path
 
 import numpy as np
@@ -41,9 +42,9 @@ class NumpyCache:
             for item, vec in zip(item, vectors):
                 item_hash = _hash_item(item)
                 if item_hash in self.hash_to_index:
-                    logger.warning(
-                        "Hash collision or duplicate item. Overwriting existing vector."
-                    )
+                    msg = f"Hash collision or duplicate item for hash {item_hash}. Overwriting existing vector."
+                    logger.warning(msg)
+                    warnings.warn(msg)
                     index = self.hash_to_index[item_hash]
                 else:
                     index = len(self.hash_to_index)
@@ -107,9 +108,9 @@ class NumpyCache:
                 f"Loaded vector dimension {self.vector_dim} from {self.dimension_file}"
             )
         else:
-            logger.warning(
-                "Dimension file not found. Vector dimension remains uninitialized."
-            )
+            msg = "Dimension file not found. Vector dimension remains uninitialized."
+            logger.warning(msg)
+            warnings.warn(msg)
 
     def save(self) -> None:
         """Persist VectorCacheMap to disk."""
@@ -151,14 +152,14 @@ class NumpyCache:
                     self.vectors = self.vectors.reshape(-1, self.vector_dim)
                     logger.info(f"Loaded vectors file with shape: {self.vectors.shape}")
                 else:
-                    logger.warning(
-                        "Vector dimension not set. Unable to load vectors file."
-                    )
+                    msg = "Vector dimension not set. Unable to load vectors file."
+                    logger.warning(msg)
+                    warnings.warn(msg)
                 logger.info(f"Loaded VectorCacheMap from {self.directory}")
             else:
-                logger.warning(
-                    "No existing files found. Initialized empty VectorCacheMap."
-                )
+                msg = "No existing files found. Initialized empty VectorCacheMap."
+                logger.warning(msg)
+                warnings.warn(msg)
         except Exception as e:
             logger.error(f"Error loading VectorCacheMap: {str(e)}")
             raise

mteb/models/model_implementations/align_models.py

@@ -105,6 +105,7 @@ class ALIGNModel(AbsEncoder):
 align_base = ModelMeta(
     loader=ALIGNModel,
     name="kakaobrain/align-base",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="e96a37facc7b1f59090ece82293226b817afd6ba",
     release_date="2023-02-24",

mteb/models/model_implementations/amazon_models.py

@@ -3,6 +3,7 @@ from mteb.models.model_meta import ModelMeta, ScoringFunction
 amazon_titan_text_embeddings_v2 = ModelMeta(
     loader=None,
     name="amazon/Titan-text-embeddings-v2",
+    model_type=["dense"],
     revision="1",
     release_date="2024-04-30",
     languages=["eng-Latn"],

mteb/models/model_implementations/andersborges.py

@@ -6,6 +6,7 @@ from mteb.models.model_meta import ModelMeta, ScoringFunction
 model2vecdk = ModelMeta(
     loader=Model2VecModel,  # type: ignore
     name="andersborges/model2vecdk",
+    model_type=["dense"],
     languages=["dan-Latn"],
     open_weights=True,
     revision="cb576c78dcc1b729e4612645f61db59929d69e61",
@@ -36,6 +37,7 @@ model2vecdk = ModelMeta(
 model2vecdk_stem = ModelMeta(
     loader=Model2VecModel,  # type: ignore
     name="andersborges/model2vecdk-stem",
+    model_type=["dense"],
     languages=["dan-Latn"],
     open_weights=True,
     revision="cb576c78dcc1b729e4612645f61db59929d69e61",

mteb/models/model_implementations/ara_models.py

@@ -4,6 +4,7 @@ from mteb.models.sentence_transformer_wrapper import sentence_transformers_loade
 arabic_triplet_matryoshka = ModelMeta(
     loader=sentence_transformers_loader,
     name="Omartificial-Intelligence-Space/Arabic-Triplet-Matryoshka-V2",
+    model_type=["dense"],
     languages=["ara-Arab"],
     open_weights=True,
     revision="ed357f222f0b6ea6670d2c9b5a1cb93950d34200",

mteb/models/model_implementations/arctic_models.py

@@ -140,6 +140,7 @@ arctic_v2_training_datasets = {
 arctic_embed_xs = ModelMeta(
     loader=sentence_transformers_loader,
     name="Snowflake/snowflake-arctic-embed-xs",
+    model_type=["dense"],
     revision="742da4f66e1823b5b4dbe6c320a1375a1fd85f9e",
     release_date="2024-07-08",  # initial commit of hf model.
     languages=["eng-Latn"],
@@ -165,6 +166,7 @@ arctic_embed_xs = ModelMeta(
 arctic_embed_s = ModelMeta(
     loader=sentence_transformers_loader,
     name="Snowflake/snowflake-arctic-embed-s",
+    model_type=["dense"],
     revision="d3c1d2d433dd0fdc8e9ca01331a5f225639e798f",
     release_date="2024-04-12",  # initial commit of hf model.
     languages=["eng-Latn"],
@@ -190,6 +192,7 @@ arctic_embed_s = ModelMeta(
 arctic_embed_m = ModelMeta(
     loader=sentence_transformers_loader,
     name="Snowflake/snowflake-arctic-embed-m",
+    model_type=["dense"],
     revision="cc17beacbac32366782584c8752220405a0f3f40",
     release_date="2024-04-12",  # initial commit of hf model.
     languages=["eng-Latn"],
@@ -215,6 +218,7 @@ arctic_embed_m_long = ModelMeta(
     loader=sentence_transformers_loader,
     loader_kwargs={"trust_remote_code": True},
     name="Snowflake/snowflake-arctic-embed-m-long",
+    model_type=["dense"],
     revision="89d0f6ab196eead40b90cb6f9fefec01a908d2d1",
     release_date="2024-04-12",  # initial commit of hf model.
     languages=["eng-Latn"],
@@ -239,6 +243,7 @@ arctic_embed_m_long = ModelMeta(
 arctic_embed_l = ModelMeta(
     loader=sentence_transformers_loader,
     name="Snowflake/snowflake-arctic-embed-l",
+    model_type=["dense"],
     revision="9a9e5834d2e89cdd8bb72b64111dde496e4fe78c",
     release_date="2024-04-12",  # initial commit of hf model.
     languages=["eng-Latn"],
@@ -268,6 +273,7 @@ arctic_embed_m_v1_5 = ModelMeta(
         },
     ),
     name="Snowflake/snowflake-arctic-embed-m-v1.5",
+    model_type=["dense"],
     revision="97eab2e17fcb7ccb8bb94d6e547898fa1a6a0f47",
     release_date="2024-07-08",  # initial commit of hf model.
     languages=["eng-Latn"],
@@ -293,6 +299,7 @@ arctic_embed_m_v2_0 = ModelMeta(
     loader=sentence_transformers_loader,
     loader_kwargs={"trust_remote_code": True},
     name="Snowflake/snowflake-arctic-embed-m-v2.0",
+    model_type=["dense"],
     revision="f2a7d59d80dfda5b1d14f096f3ce88bb6bf9ebdc",
     release_date="2024-12-04",  # initial commit of hf model.
     languages=LANGUAGES_V2_0,
@@ -317,6 +324,7 @@ arctic_embed_m_v2_0 = ModelMeta(
 arctic_embed_l_v2_0 = ModelMeta(
     loader=sentence_transformers_loader,
     name="Snowflake/snowflake-arctic-embed-l-v2.0",
+    model_type=["dense"],
     revision="edc2df7b6c25794b340229ca082e7c78782e6374",
     release_date="2024-12-04",  # initial commit of hf model.
     languages=LANGUAGES_V2_0,

mteb/models/model_implementations/b1ade_models.py

@@ -10,6 +10,7 @@ b1ade_training_data = {
 b1ade_embed = ModelMeta(
     loader=sentence_transformers_loader,
     name="w601sxs/b1ade-embed",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="3bdac13927fdc888b903db93b2ffdbd90b295a69",
     open_weights=True,

mteb/models/model_implementations/bedrock_models.py

@@ -155,6 +155,7 @@ class BedrockModel(AbsEncoder):
 
 amazon_titan_embed_text_v1 = ModelMeta(
     name="bedrock/amazon-titan-embed-text-v1",
+    model_type=["dense"],
     revision="1",
     release_date="2023-09-27",
     languages=None,  # not specified
@@ -181,6 +182,7 @@ amazon_titan_embed_text_v1 = ModelMeta(
 
 amazon_titan_embed_text_v2 = ModelMeta(
     name="bedrock/amazon-titan-embed-text-v2",
+    model_type=["dense"],
     revision="1",
     release_date="2024-04-30",
     languages=None,  # not specified
@@ -216,6 +218,7 @@ cohere_embed_english_v3 = ModelMeta(
         model_prompts=cohere_model_prompts,
     ),
     name="bedrock/cohere-embed-english-v3",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=False,
     reference="https://cohere.com/blog/introducing-embed-v3",
@@ -243,6 +246,7 @@ cohere_embed_multilingual_v3 = ModelMeta(
         model_prompts=cohere_model_prompts,
     ),
     name="bedrock/cohere-embed-multilingual-v3",
+    model_type=["dense"],
     languages=cohere_supported_languages,
     open_weights=False,
     reference="https://cohere.com/blog/introducing-embed-v3",