fix: Add cadet-embed-base-v1

mteb/models/cadet_models.py

@@ -0,0 +1,60 @@
+from __future__ import annotations
+
+from functools import partial
+
+from mteb.model_meta import ModelMeta, sentence_transformers_loader
+from mteb.models.bge_models import bge_m3_training_data
+
+cadet_training_data = {
+    # we train with the corpora of FEVER, MSMARCO, and DBPEDIA. We only train with synthetic generated queries.
+    # However, we do use queries from MSMARCO as examples for synthetic query generation.
+    "MSMARCO": ["train"],
+    "DBPedia": [],
+    # We distill from RankT5 which trains using MSMARCO + NQ.
+    # source: https://arxiv.org/pdf/2210.10634
+    "NQ": ["train"],
+    # We also distill from bge-rerankerv2.5-gemma2-lightweight, which utilizes the BGE-M3 dataset for training, along with Arguana, HotpotQA, and FEVER.
+    # source: https://arxiv.org/pdf/2409.15700
+    "FEVER": ["train"],
+    "HotpotQA": ["train"],
+    "ArguAna": ["train"],
+}
+
+for k, v in bge_m3_training_data.items():
+    cadet_training_data.setdefault(k, []).extend(v)
+# deduplicate
+cadet_training_data = {
+    k: list(dict.fromkeys(v)) for k, v in cadet_training_data.items()
+}
+
+
+cadet_embed = ModelMeta(
+    loader=partial(
+        sentence_transformers_loader,
+        model_name="manveertamber/cadet-embed-base-v1",
+        revision="8056d118be37a566f20972a5f35cda815f6bc47e",
+        model_prompts={
+            "query": "query: ",
+            "passage": "passage: ",
+        },
+    ),
+    name="manveertamber/cadet-embed-base-v1",
+    languages=["eng-Latn"],
+    revision="8056d118be37a566f20972a5f35cda815f6bc47e",
+    open_weights=True,
+    release_date="2025-05-11",
+    n_parameters=109_000_000,
+    memory_usage_mb=418,
+    embed_dim=768,
+    license="apache-2.0",
+    max_tokens=512,
+    reference="https://huggingface.co/manveertamber/cadet-embed-base-v1",
+    similarity_fn_name="cosine",
+    framework=["Sentence Transformers", "PyTorch"],
+    use_instructions=True,
+    public_training_code="https://github.com/manveertamber/cadet-dense-retrieval",
+    # we provide the code to generate the training data
+    public_training_data="https://github.com/manveertamber/cadet-dense-retrieval",
+    training_datasets=cadet_training_data,
+    adapted_from="intfloat/e5-base-unsupervised",
+)

mteb/models/overview.py

@@ -21,6 +21,7 @@
     blip2_models,
     blip_models,
     bm25,
+    cadet_models,
     cde_models,
     clip_models,
     codesage_models,
@@ -96,6 +97,7 @@
     blip2_models,
     blip_models,
     bm25,
+    cadet_models,
     clip_models,
     codesage_models,
     cde_models,