updated dataset

app.py

@@ -12,14 +12,14 @@ LLM_PERF_LEADERBOARD_REPO = "optimum/llm-perf-leaderboard"
 LLM_PERF_DATASET_REPO = "optimum/llm-perf-dataset"
 OPTIMUM_TOKEN = os.environ.get("OPTIMUM_TOKEN")
 
-OLD_COLUMNS = ["model", "backend.name", "backend.torch_dtype",
-               "generate.latency(s)", "generate.throughput(tokens/s)"]
-
-NEW_COLUMNS = ["Model", "Backend 🏭", "Load Datatype",
-               "Latency (s) ⬇️", "Throughput (tokens/s) ⬆️"]
-
+COLUMNS_MAPPING = {
+    "model": "Model 🤗",
+    "backend.name": "Backend 🏭",
+    "backend.torch_dtype": "Load Datatype 📥",
+    "generate.latency(s)": "Latency (s) ⬇️",
+    "generate.throughput(tokens/s)": "Throughput (tokens/s) ⬆️",
+}
 COLUMNS_DATATYPES = ["markdown", "str", "str", "number", "number"]
-
 SORTING_COLUMN = ["Throughput (tokens/s) ⬆️"]
 
 
@@ -31,20 +31,15 @@ def get_benchmark_df(benchmark):
         llm_perf_dataset_repo.git_pull()
 
     # load
-    df = pd.read_csv(
-        f"./llm-perf-dataset/reports/{benchmark}/inference_report.csv")
-
+    df = pd.read_csv(f"llm-perf-dataset/reports/{benchmark}/inference_report.csv")
     # preprocess
     df["model"] = df["model"].apply(make_clickable_model)
-
     # filter
-    df = df[OLD_COLUMNS]
-
+    df = df[COLUMNS_MAPPING.keys()]
     # rename
     df.rename(columns={
-        df_col: rename_col for df_col, rename_col in zip(OLD_COLUMNS, NEW_COLUMNS)
+        df_col: rename_col for df_col, rename_col in COLUMNS_MAPPING.items()
     }, inplace=True)
-
     # sort
     df.sort_values(by=SORTING_COLUMN, ascending=False, inplace=True)
 
@@ -72,7 +67,7 @@ with demo:
             leaderboard_table_lite = gr.components.Dataframe(
                 value=single_A100_df,
                 datatype=COLUMNS_DATATYPES,
-                headers=NEW_COLUMNS,
+                headers=COLUMNS_MAPPING.values(),
                 elem_id="1xA100-table",
             )
 
@@ -87,7 +82,7 @@ with demo:
             leaderboard_table_full = gr.components.Dataframe(
                 value=multi_A100_df,
                 datatype=COLUMNS_DATATYPES,
-                headers=NEW_COLUMNS,
+                headers=COLUMNS_MAPPING.values(),
                 elem_id="4xA100-table",
             )
 
@@ -100,6 +95,7 @@ with demo:
                     elem_id="citation-button",
                 ).style(show_copy_button=True)
 
+
 # Restart space every hour
 scheduler = BackgroundScheduler()
 scheduler.add_job(restart_space, "interval", seconds=3600,