style update

app.py

@@ -15,7 +15,7 @@ H4_TOKEN = os.environ.get("H4_TOKEN", None)
 LMEH_REPO = "HuggingFaceH4/lmeh_evaluations"
 IS_PUBLIC = bool(os.environ.get("IS_PUBLIC", None))
 
-repo=None
+repo = None
 if H4_TOKEN:
     print("pulling repo")
     # try:
@@ -24,7 +24,10 @@ if H4_TOKEN:
     #     pass
 
     repo = Repository(
-        local_dir="./evals/", clone_from=LMEH_REPO, use_auth_token=H4_TOKEN, repo_type="dataset"
+        local_dir="./evals/",
+        clone_from=LMEH_REPO,
+        use_auth_token=H4_TOKEN,
+        repo_type="dataset",
     )
     repo.git_pull()
 
@@ -47,15 +50,33 @@ def load_results(model, benchmark, metric):
     return mean_acc, data["config"]["model_args"]
 
 
-COLS = ["Model", "Revision", "Average ⬆️", "ARC (25-shot) ⬆️", "HellaSwag (10-shot) ⬆️", "MMLU (5-shot) ⬆️", "TruthfulQA (0-shot) ⬆️"]
-TYPES = ["markdown","str", "number", "number", "number", "number", "number", ]
+COLS = [
+    "Model",
+    "Revision",
+    "Average ⬆️",
+    "ARC (25-shot) ⬆️",
+    "HellaSwag (10-shot) ⬆️",
+    "MMLU (5-shot) ⬆️",
+    "TruthfulQA (0-shot) ⬆️",
+]
+TYPES = [
+    "markdown",
+    "str",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+]
 
 if not IS_PUBLIC:
     COLS.insert(2, "8bit")
     TYPES.insert(2, "bool")
 
 EVAL_COLS = ["model", "revision", "private", "8bit_eval", "is_delta_weight", "status"]
-EVAL_TYPES = ["markdown","str", "bool", "bool", "bool", "str"]
+EVAL_TYPES = ["markdown", "str", "bool", "bool", "bool", "str"]
+
+
 def get_leaderboard():
     if repo:
         print("pulling changes")
@@ -65,44 +86,49 @@ def get_leaderboard():
 
     if not IS_PUBLIC:
         gpt4_values = {
-            "Model":f'<a target="_blank" href=https://arxiv.org/abs/2303.08774 style="color: blue; text-decoration: underline;text-decoration-style: dotted;">gpt4</a>',
-            "Revision":"tech report",
-            "8bit":None,
-            "Average ⬆️":84.3,
-            "ARC (25-shot) ⬆️":96.3,
-            "HellaSwag (10-shot) ⬆️":95.3,
-            "MMLU (5-shot) ⬆️":86.4,
-            "TruthfulQA (0-shot) ⬆️":59.0,
+            "Model": f'<a target="_blank" href=https://arxiv.org/abs/2303.08774 style="color: blue; text-decoration: underline;text-decoration-style: dotted;">gpt4</a>',
+            "Revision": "tech report",
+            "8bit": None,
+            "Average ⬆️": 84.3,
+            "ARC (25-shot) ⬆️": 96.3,
+            "HellaSwag (10-shot) ⬆️": 95.3,
+            "MMLU (5-shot) ⬆️": 86.4,
+            "TruthfulQA (0-shot) ⬆️": 59.0,
         }
         all_data.append(gpt4_values)
         gpt35_values = {
-            "Model":f'<a target="_blank" href=https://arxiv.org/abs/2303.08774 style="color: blue; text-decoration: underline;text-decoration-style: dotted;">gpt3.5</a>',
-            "Revision":"tech report",
-            "8bit":None,
-            "Average ⬆️":71.9,
-            "ARC (25-shot) ⬆️":85.2,
-            "HellaSwag (10-shot) ⬆️":85.5,
-            "MMLU (5-shot) ⬆️":70.0,
-            "TruthfulQA (0-shot) ⬆️":47.0,
+            "Model": f'<a target="_blank" href=https://arxiv.org/abs/2303.08774 style="color: blue; text-decoration: underline;text-decoration-style: dotted;">gpt3.5</a>',
+            "Revision": "tech report",
+            "8bit": None,
+            "Average ⬆️": 71.9,
+            "ARC (25-shot) ⬆️": 85.2,
+            "HellaSwag (10-shot) ⬆️": 85.5,
+            "MMLU (5-shot) ⬆️": 70.0,
+            "TruthfulQA (0-shot) ⬆️": 47.0,
         }
         all_data.append(gpt35_values)
 
     dataframe = pd.DataFrame.from_records(all_data)
-    dataframe = dataframe.sort_values(by=['Average ⬆️'], ascending=False)
+    dataframe = dataframe.sort_values(by=["Average ⬆️"], ascending=False)
     print(dataframe)
     dataframe = dataframe[COLS]
     return dataframe
 
+
 def get_eval_table():
     if repo:
         print("pulling changes for eval")
         repo.git_pull()
-    entries = [entry for entry in os.listdir("evals/eval_requests") if not entry.startswith('.')]
+    entries = [
+        entry
+        for entry in os.listdir("evals/eval_requests")
+        if not entry.startswith(".")
+    ]
     all_evals = []
 
     for entry in entries:
         print(entry)
-        if ".json"in entry:
+        if ".json" in entry:
             file_path = os.path.join("evals/eval_requests", entry)
             with open(file_path) as fp:
                 data = json.load(fp)
@@ -111,21 +137,23 @@ def get_eval_table():
             data["model"] = make_clickable_model(data["model"])
             data["revision"] = data.get("revision", "main")
 
-
             all_evals.append(data)
         else:
             # this is a folder
-            sub_entries = [e for e in os.listdir(f"evals/eval_requests/{entry}") if not e.startswith('.')]
+            sub_entries = [
+                e
+                for e in os.listdir(f"evals/eval_requests/{entry}")
+                if not e.startswith(".")
+            ]
             for sub_entry in sub_entries:
                 file_path = os.path.join("evals/eval_requests", entry, sub_entry)
                 with open(file_path) as fp:
                     data = json.load(fp)
 
-                #data["# params"] = get_n_params(data["model"])
+                # data["# params"] = get_n_params(data["model"])
                 data["model"] = make_clickable_model(data["model"])
                 all_evals.append(data)
 
-
     dataframe = pd.DataFrame.from_records(all_evals)
     return dataframe[EVAL_COLS]
 
@@ -133,6 +161,7 @@ def get_eval_table():
 leaderboard = get_leaderboard()
 eval_queue = get_eval_table()
 
+
 def is_model_on_hub(model_name, revision) -> bool:
     try:
         config = AutoConfig.from_pretrained(model_name, revision=revision)
@@ -144,7 +173,6 @@ def is_model_on_hub(model_name, revision) -> bool:
         return False
 
 
-
 def add_new_eval(
     model: str,
     base_model: str,
@@ -157,15 +185,15 @@ def add_new_eval(
     if revision == "":
         revision = "main"
     if is_delta_weight and not is_model_on_hub(base_model, revision):
-        error_message = f"Base model \"{base_model}\" was not found on hub!"
+        error_message = f'Base model "{base_model}" was not found on hub!'
         print(error_message)
         return f"<p style='color: red; font-size: 18px; text-align: center;'>{error_message}</p>"
 
     if not is_model_on_hub(model, revision):
-        error_message = f"Model \"{model}\"was not found on hub!"
+        error_message = f'Model "{model}"was not found on hub!'
         print(error_message)
         return f"<p style='color: red; font-size: 18px; text-align: center;'>{error_message}</p>"
-    
+
     print("adding new eval")
 
     eval_entry = {
@@ -200,7 +228,7 @@ def add_new_eval(
         token=H4_TOKEN,
         repo_type="dataset",
     )
-    
+
     success_message = "Your request has been submitted to the evaluation queue!"
     return f"<p style='color: green; font-size: 18px; text-align: center;'>{success_message}</p>"
 
@@ -209,11 +237,11 @@ def refresh():
     return get_leaderboard(), get_eval_table()
 
 
-
 block = gr.Blocks()
 with block:
     with gr.Row():
-        gr.Markdown(f"""
+        gr.Markdown(
+            f"""
 # 🤗 Open LLM Leaderboard
 <font size="4">With the plethora of large language models (LLMs) and chatbots being released week upon week, often with grandiose claims of their performance, it can be hard to filter out the genuine progress that is being made by the open-source community and which model is the current state of the art. The 🤗 Open LLM Leaderboard aims to track, rank and evaluate LLMs and chatbots as they are released. We evaluate models on 4 key benchmarks from the <a href="https://github.com/EleutherAI/lm-evaluation-harness" target="_blank">  Eleuther AI Language Model Evaluation Harness </a>, a unified framework to test generative language models on a large number of different evaluation tasks. A key advantage of this leaderboard is that anyone from the community can submit a model for automated evaluation on the 🤗 GPU cluster, as long as it is a 🤗 Transformers model with weights on the Hub. We also support evaluation of models with delta-weights for non-commercial licensed models, such as LLaMa.
 
@@ -224,27 +252,32 @@ Evaluation is performed against 4 popular benchmarks:
 - <a href="https://arxiv.org/abs/2109.07958" target="_blank">  TruthfulQA </a> (0-shot) - a benchmark to measure whether a language model is truthful in generating answers to questions.
 
 We chose these benchmarks as they test a variety of reasoning and general knowledge across a wide variety of fields in 0-shot and few-shot settings. </font>
-        """)
+        """
+        )
 
     with gr.Row():
-        leaderboard_table = gr.components.Dataframe(value=leaderboard, headers=COLS,
-                                                    datatype=TYPES, max_rows=5)
-
-
+        leaderboard_table = gr.components.Dataframe(
+            value=leaderboard, headers=COLS, datatype=TYPES, max_rows=5
+        )
 
     with gr.Row():
-        gr.Markdown(f"""
+        gr.Markdown(
+            f"""
     # Evaluation Queue for the 🤗 Open LLM Leaderboard, these models will be automatically evaluated on the 🤗 cluster
 
-    """)
+    """
+        )
     with gr.Accordion("Evaluation Queue", open=False):
         with gr.Row():
-            eval_table = gr.components.Dataframe(value=eval_queue, headers=EVAL_COLS,
-                                                        datatype=EVAL_TYPES, max_rows=5)
+            eval_table = gr.components.Dataframe(
+                value=eval_queue, headers=EVAL_COLS, datatype=EVAL_TYPES, max_rows=5
+            )
 
     with gr.Row():
         refresh_button = gr.Button("Refresh")
-        refresh_button.click(refresh, inputs=[], outputs=[leaderboard_table, eval_table])
+        refresh_button.click(
+            refresh, inputs=[], outputs=[leaderboard_table, eval_table]
+        )
 
     with gr.Accordion("Submit a new model for evaluation"):
         with gr.Row():
@@ -253,7 +286,9 @@ We chose these benchmarks as they test a variety of reasoning and general knowle
                 revision_name_textbox = gr.Textbox(label="revision", placeholder="main")
 
             with gr.Column():
-                is_8bit_toggle = gr.Checkbox(False, label="8 bit eval", visible=not IS_PUBLIC)
+                is_8bit_toggle = gr.Checkbox(
+                    False, label="8 bit eval", visible=not IS_PUBLIC
+                )
                 private = gr.Checkbox(False, label="Private", visible=not IS_PUBLIC)
                 is_delta_weight = gr.Checkbox(False, label="Delta weights")
                 base_model_name_textbox = gr.Textbox(label="base model (for delta)")
@@ -263,18 +298,17 @@ We chose these benchmarks as they test a variety of reasoning and general knowle
         with gr.Row():
             submission_result = gr.Markdown()
             submit_button.click(
-                            add_new_eval,
-                            [
-                                model_name_textbox,
-                                base_model_name_textbox,
-                                revision_name_textbox,
-                                is_8bit_toggle,
-                                private,
-                                is_delta_weight,
-                            ],
-                            submission_result
-                        )
-
+                add_new_eval,
+                [
+                    model_name_textbox,
+                    base_model_name_textbox,
+                    revision_name_textbox,
+                    is_8bit_toggle,
+                    private,
+                    is_delta_weight,
+                ],
+                submission_result,
+            )
 
     block.load(refresh, inputs=[], outputs=[leaderboard_table, eval_table])
 block.launch()