model: Add NIFE models

mteb/models/model_implementations/stephantulkens_models.py

@@ -0,0 +1,81 @@
+from mteb.models.model_meta import ModelMeta, ScoringFunction
+from mteb.models.sentence_transformer_wrapper import sentence_transformers_loader
+
+TRAINING_DATA = {
+    "MSMARCO",
+    "MIRACLRetrieval",
+    "LoTTE",
+    "TriviaQA",
+    "GooAQ",
+    "MrTidyRetrieval",
+}
+
+CITAITON = """@software{Tulkens2025pyNIFE,
+  author       = {St\'{e}phan Tulkens},
+  title        = {pyNIFE: nearly inference free embeddings in python},
+  year         = {2025},
+  publisher    = {Zenodo},
+  doi          = {10.5281/zenodo.17512919},
+  url          = {https://github.com/stephantul/pynife},
+  license      = {MIT},
+}"""
+
+ModelMeta(
+    loader=sentence_transformers_loader,  # type: ignore[call-arg]
+    loader_kwargs={},
+    name="stephantulkens/NIFE-gte-modernbert-base",
+    revision="9b85ba93d5a7729db11dfea90d20a5f392a88747",
+    release_date="2025-10-30",
+    languages=["eng-Latn"],  # assumed from the original model
+    n_parameters=76802304,  # TODO: what do we do for routers? Both models I assume? - this is for the query router / student model
+    n_active_parameters_override=None,  # TODO: not sure how to count this for routers - WDYT?
+    n_embedding_parameters=76802304,  # this is for the query router / student model
+    memory_usage_mb=293.0,
+    max_tokens=8192.0,
+    embed_dim=768,
+    license="mit",
+    open_weights=True,
+    public_training_code=None,  # no reproducible training code, but likely uses https://pypi.org/project/pynife/
+    public_training_data="https://huggingface.co/collections/stephantulkens/gte-modernbert-embedpress",  # + its a distilaltion of a model
+    framework=["PyTorch", "safetensors", "Sentence Transformers"],
+    reference="https://huggingface.co/stephantulkens/NIFE-gte-modernbert-base",
+    similarity_fn_name=ScoringFunction.COSINE,
+    use_instructions=False,  # assumed
+    training_datasets=TRAINING_DATA,
+    superseded_by=None,
+    modalities=["text"],
+    model_type=["dense"],  # TODO: is router a model type?
+    citation=CITAITON,
+    contacts=None,
+    adapted_from="Alibaba-NLP/gte-modernbert-base",
+)
+
+ModelMeta(
+    loader=sentence_transformers_loader,  # type: ignore[call-arg]
+    loader_kwargs={},
+    name="stephantulkens/NIFE-mxbai-embed-large-v1",
+    revision="970a0151f3faeddbab5c68be38d52bbc5ea719ab",
+    release_date="2025-11-03",
+    languages=["eng-Latn"],  # assumed from the original model
+    n_parameters=110553088,  # TODO: what do we do for routers? Both models I assume? - this is for the query router / student model
+    n_active_parameters_override=None,  # TODO: not sure how to count this for routers - WDYT?
+    n_embedding_parameters=110553088,  # this is for the query router / student model
+    memory_usage_mb=422.0,
+    max_tokens=512,  # assumed from the original model
+    embed_dim=1024,
+    license="mit",
+    open_weights=True,
+    public_training_code=None,  # no reproducible training code, but likely uses https://pypi.org/project/pynife/
+    public_training_data=None,  # its a distilaltion of a model
+    framework=["PyTorch", "safetensors", "Sentence Transformers"],
+    reference="https://huggingface.co/stephantulkens/NIFE-mxbai-embed-large-v1",
+    similarity_fn_name=ScoringFunction.COSINE,
+    use_instructions=False,  # assumed
+    training_datasets=TRAINING_DATA,
+    adapted_from="mixedbread-ai/mxbai-embed-large-v1",
+    superseded_by=None,
+    modalities=["text"],
+    model_type=["dense"],  # TODO: is router a model type?
+    citation=CITAITON,
+    contacts=None,
+)