add bi-encoder button

app.py

@@ -44,12 +44,12 @@ TASK_DESCRIPTIONS["Overall"] = "Overall performance across MTEB tasks."
 SENTENCE_TRANSFORMERS_COMPATIBLE_MODELS = {k for k,v in MODEL_META["model_meta"].items() if v.get("is_sentence_transformers_compatible", False)}
 MODELS_TO_SKIP = MODEL_META["models_to_skip"]
 CROSS_ENCODERS = MODEL_META["cross_encoders"]
+BI_ENCODERS = [k for k, _ in MODEL_META["model_meta"].items() if k not in CROSS_ENCODERS + ["bm25"]]
 
 PROPRIETARY_MODELS = {
     make_clickable_model(model, link=EXTERNAL_MODEL_TO_LINK.get(model, f"https://huggingface.co/spaces/{REPO_ID}"))
     for model in PROPRIETARY_MODELS
 }
-
 SENTENCE_TRANSFORMERS_COMPATIBLE_MODELS = {
     make_clickable_model(model, link=EXTERNAL_MODEL_TO_LINK.get(model, f"https://huggingface.co/spaces/{REPO_ID}"))
     for model in SENTENCE_TRANSFORMERS_COMPATIBLE_MODELS
@@ -58,6 +58,11 @@ CROSS_ENCODERS = {
     make_clickable_model(model, link=EXTERNAL_MODEL_TO_LINK.get(model, f"https://huggingface.co/spaces/{REPO_ID}"))
     for model in CROSS_ENCODERS
 }
+BI_ENCODERS = {
+    make_clickable_model(model, link=EXTERNAL_MODEL_TO_LINK.get(model, f"https://huggingface.co/spaces/{REPO_ID}"))
+    for model in BI_ENCODERS
+}
+
 
 TASK_TO_TASK_TYPE = {task_category: [] for task_category in TASKS}
 for board_config in BOARDS_CONFIG.values():
@@ -482,7 +487,8 @@ MODEL_TYPES = [
     "Open",
     "Proprietary",
     "Sentence Transformers",
-    "Cross Encoders"
+    "Cross-Encoders",
+    "Bi-Encoders"
 ]
 
 def filter_data(search_query, model_types, model_sizes, *full_dataframes):
@@ -506,8 +512,10 @@ def filter_data(search_query, model_types, model_sizes, *full_dataframes):
                     masks.append(df["Model"].isin(PROPRIETARY_MODELS))
                 elif model_type == "Sentence Transformers":
                     masks.append(df["Model"].isin(SENTENCE_TRANSFORMERS_COMPATIBLE_MODELS))
-                elif model_type == "Cross Encoders":
+                elif model_type == "Cross-Encoders":
                     masks.append(df["Model"].isin(CROSS_ENCODERS))
+                elif model_type == "Bi-Encoders":
+                    masks.append(df["Model"].isin(BI_ENCODERS))
             if masks:
                 df = df[reduce(lambda a, b: a | b, masks)]
             else:

model_meta.yaml

@@ -1306,4 +1306,7 @@ cross_encoders:
 - flan-t5-large
 - monobert-large-msmarco
 - monot5-3b-msmarco-10k
-- monot5-base-msmarco-10
+- monot5-base-msmarco-10k
+- llama-2-7b-chat
+- mistral-7b-instruct-v0.2
+- tart-full-flan-t5-xl