visibility

app.py

@@ -1,7 +1,7 @@
 import gradio as gr  # type: ignore
 import plotly.express as px  # type: ignore
 
-from backend.data import load_cot_data
+from backend.data import load_cot_data, is_visible_model
 from backend.envs import API, REPO_ID, TOKEN
 
 logo1_url = "https://raw.githubusercontent.com/logikon-ai/cot-eval/main/assets/AI2_Logo_Square.png"
@@ -38,6 +38,7 @@ def plot_evals_init(model_id, plotly_mode, request: gr.Request):
 def plot_evals(model_id, plotly_mode):
     df = df_cot_err.copy()
     df["selected"] = df_cot_err.model.apply(lambda x: "selected" if x==model_id else "-")
+    df["visible"] = df_cot_err.model.apply(is_visible_model) | df.selected.eq("selected")
     #df.sort_values(["selected", "model"], inplace=True, ascending=True)  # has currently no effect with px.scatter
     template = "plotly_dark" if plotly_mode=="dark" else "plotly" 
     fig = px.scatter(df, x="base accuracy", y="marginal acc. gain", color="selected", symbol="model",
@@ -48,6 +49,11 @@ def plot_evals(model_id, plotly_mode):
                     error_y="acc_gain-err", hover_data=['model', "cot accuracy"],
                     width=1200, height=700)
 
+    fig.update_traces(
+        visible="legendonly",
+        selector=dict(visible=False)
+    )
+
     fig.update_layout(
         title={"automargin": True},
     )

backend/data.py

@@ -9,6 +9,24 @@ from backend.envs import EVAL_DATASET, TRACES_DATASET, TOKEN, EVAL_RESULTS_PATH
 
 
 SUBSETS = ["base","cot","orig"]
+INITIALLY_VISIBLE_MODELS = [
+    "01-ai/Yi-34B-Chat",
+    "Qwen/Qwen2-72B-Instruct",
+    "allenai/tulu-2-dpo-70b",
+    "google/gemma-2-2b-it",
+    "internlm/internlm2-chat-20b",
+    "meta-llama/Llama-2-70b-chat-hf",
+    "meta-llama/Meta-Llama-3.1-70B-Instruct",
+    "meta-llama/Meta-Llama-3.1-8B-Instruct",
+    "microsoft/Phi-3-mini-4k-instruct",
+    "microsoft/Phi-3.5-MoE-instruct",
+    "mistralai/Mistral-7B-Instruct-v0.2",
+    "mistralai/Mixtral-8x7B-Instruct-v0.1",
+]
+
+
+def is_visible_model(model: str) -> bool:
+    return model in INITIALLY_VISIBLE_MODELS
 
 
 def load_cot_data():