Merge branch 'main' of https://github.com/KennethEnevoldsen/scandinav…

…ian-embedding-benchmark

docs/index.md

@@ -6,13 +6,13 @@ hide:
 
 # Scandinavian Embedding Benchmark
 
-This is the documentation for the Scandinavian Embedding Benchmark. This benchmark is intended to evaluate the sentence/documents embeddings of large language models.
+This is the documentation for the Scandinavian Embedding Benchmark. This benchmark is intended to evaluate the sentence/document embeddings of large language models.
 
 Intended uses for this benchmark:
 
 - Evaluating document embeddings of Scandinavian language models
 - Evaluating document embeddings for multilingual models on Scandinavian languages
-- Allow ranking of competing Scandinavian and multilingual models using no more compute that what a consumer laptop can provide 
+- Allow ranking of competing Scandinavian and multilingual models using no more compute than what a consumer laptop can provide 
 
 
 === "All"
@@ -34,9 +34,9 @@ Intended uses for this benchmark:
 
 ## Comparison to other benchmarks
 
-If you use this benchmark for a relative ranking of language models you should also take a look at [ScandEval](https://scandeval.github.io), which as opposed the this benchmark fully fine-tunes the models. It also includes structured predictions tasks such as named entity recognition. Many of the tasks in this embeddings benchmark is also included in ScandEval. A notable difference between the ScandEval and this benchmark is that it does not include machine translated tasks.
+If you use this benchmark for a relative ranking of language models you should also look at [ScandEval](https://scandeval.github.io), which as opposed to this benchmark fully fine-tunes the models. It also includes structured prediction tasks such as named entity recognition. Many of the tasks in this embedding benchmark are also included in ScandEval. A notable difference between ScandEval and this benchmark is that this one does not include machine-translated tasks.
 
-The tasks within this benchmark is also included in the [MTEB](https://huggingface.co/spaces/mteb/leaderboard) leaderboard, though the aggregations methods very slightly. The MTEB is primarily an English embedding benchmark, with a few multilingual tasks along with a few additional languages. As a part of this project the tasks was also added to the MTEB leaderboard.
+The tasks within this benchmark are also included in the [MTEB](https://huggingface.co/spaces/mteb/leaderboard) leaderboard, though the aggregation methods very slightly. MTEB is primarily an English embedding benchmark, with a few multilingual tasks and additional languages. As a part of this project, the tasks were also added to the MTEB leaderboard.
 
 
 

src/seb/seb_models.py

@@ -41,6 +41,19 @@ def create_all_mini_lm_l6_v2() -> SebModel:
         meta=meta,
     )
 
+@models.register("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
+def create_multilingual_mini_lm_l12_v2() -> SebModel:
+    hf_name = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
+    meta = ModelMeta(
+        name=hf_name.split("/")[-1],
+        huggingface_name=hf_name,
+        reference=f"https://huggingface.co/{hf_name}",
+        languages=[],
+    )
+    return SebModel(
+        loader=partial(get_sentence_transformer, model_name=hf_name),  # type: ignore
+        meta=meta,
+    )
 
 @models.register("KBLab/sentence-bert-swedish-cased")
 def create_sentence_swedish_cased() -> SebModel: