updated scores

app.py

@@ -33,7 +33,7 @@ COLUMNS_MAPPING = {
     #
     "generate.throughput(tokens/s)": "Throughput (tokens/s) ⬆️",
     "forward.peak_memory(MB)": "Peak Memory (MB) ⬇️",
-    "average": "Average Open LLM Score ⬆️",
+    "score": "Average Open LLM Score ⬆️",
     #
     "num_parameters": "#️⃣ Parameters 📏",
 }
@@ -61,7 +61,7 @@ def get_benchmark_df(benchmark="1xA100-80GB"):
 
     # load
     bench_df = pd.read_csv(f"./llm-perf-dataset/reports/{benchmark}.csv")
-    scores_df = pd.read_csv(f"./llm-perf-dataset/reports/additional_data.csv")
+    scores_df = pd.read_csv(f"./llm-perf-dataset/reports/open-llm-leaderboard.csv")
     bench_df = bench_df.merge(scores_df, on="model", how="left")
 
     bench_df["optimizations"] = bench_df[
@@ -99,7 +99,7 @@ def get_benchmark_plot(bench_df):
     fig = px.scatter(
         bench_df,
         x="generate.latency(s)",
-        y="average",
+        y="score",
         color="model_type",
         symbol="backend.name",
         size="forward.peak_memory(MB)",
@@ -175,7 +175,7 @@ def filter_query(
             if len(optimizations) > 0
             else True
         )
-        & (raw_df["average"] >= score)
+        & (raw_df["score"] >= score)
         & (raw_df["forward.peak_memory(MB)"] <= memory)
     ]
 

src/utils.py

@@ -1,4 +1,3 @@
-import re
 from huggingface_hub import HfApi, Repository