Update app.py

app.py

@@ -1,21 +1,21 @@
+import os
 import json
-
 import requests
 
-from datasets import load_dataset
-
 import gradio as gr
-from apscheduler.schedulers.background import BackgroundScheduler
-
-
-from huggingface_hub import HfApi, hf_hub_download
-from huggingface_hub.repocard import metadata_load
 import pandas as pd
+from huggingface_hub import HfApi, hf_hub_download, snapshot_download
+from huggingface_hub.repocard import metadata_load
+from apscheduler.schedulers.background import BackgroundScheduler
 
 from utils import *
 
+DATASET_REPO_URL = "https://huggingface.co/datasets/huggingface-projects/drlc-leaderboard-data"
+DATASET_REPO_ID = "huggingface-projects/drlc-leaderboard-data"
+HF_TOKEN = os.environ.get("HF_TOKEN")
 
 block = gr.Blocks()
+api = HfApi(token=HF_TOKEN)
 
 # Containing the data
 rl_envs = [
@@ -135,8 +135,6 @@ rl_envs = [
 }
 ]
 
-
-
 def get_metadata(model_id):
     try:
         readme_path = hf_hub_download(model_id, filename="README.md")
@@ -181,13 +179,11 @@ def get_model_ids(rl_env):
     api = HfApi()
     models = api.list_models(filter=rl_env)
     model_ids = [x.modelId for x in models]
-    #print(model_ids)
     return model_ids
 
-def get_model_dataframe(rl_env):
+def update_leaderboard_dataset(rl_env):
     # Get model ids associated with rl_env
     model_ids = get_model_ids(rl_env)
-    #print(model_ids)
     data = []
     for model_id in model_ids:
         """
@@ -200,8 +196,8 @@ def get_model_dataframe(rl_env):
             continue
         user_id = model_id.split('/')[0]
         row = {}
-        row["User"] = make_clickable_user(user_id)
-        row["Model"] = make_clickable_model(model_id)
+        row["User"] = user_id
+        row["Model"] = model_id
         accuracy = parse_metrics_accuracy(meta)
         mean_reward, std_reward = parse_rewards(accuracy)
         mean_reward = mean_reward if not pd.isna(mean_reward) else 0
@@ -210,14 +206,53 @@ def get_model_dataframe(rl_env):
         row["Mean Reward"] = mean_reward
         row["Std Reward"] = std_reward
         data.append(row)
-    print("DATA", data)
+
     ranked_dataframe = rank_dataframe(pd.DataFrame.from_records(data))
-    print("RANKED", ranked_dataframe)
+    new_history = ranked_dataframe
+    filename = rl_env + ".csv"
+    new_history.to_csv(filename, index=False)
+    csv_path = os.path.abspath(filename)
+
+    api.upload_file(
+    path_or_fileobj= csv_path,
+    path_in_repo= filename,
+    repo_id="huggingface-projects/drlc-leaderboard-data",
+    repo_type="dataset",
+    commit_message="Update dataset")
+
     return ranked_dataframe
-    
+
+def download_leaderboard_dataset():
+    path = snapshot_download(repo_id=DATASET_REPO_ID, repo_type="dataset")
+    return path
+
+def get_data(rl_env, path) -> pd.DataFrame:
+    """
+    Get data from rl_env
+    :return: data as a pandas DataFrame
+    """
+    csv_path = path + "/" + rl_env + ".csv"
+    data = pd.read_csv(csv_path)
+
+    for index, row in data.iterrows():
+        user_id = row["User"]
+        data.loc[index, "User"] = make_clickable_user(user_id)
+        model_id = row["Model"]
+        data.loc[index, "Model"] = make_clickable_model(model_id)
+        
+    return data
+
+def get_data_no_html(rl_env, path) -> pd.DataFrame:
+    """
+    Get data from rl_env
+    :return: data as a pandas DataFrame
+    """
+    csv_path = path + "/" + rl_env + ".csv"
+    data = pd.read_csv(csv_path)
+
+    return data
     
 def rank_dataframe(dataframe):
-    #print("DATAFRAME", dataframe)
     dataframe = dataframe.sort_values(by=['Results'], ascending=False)
     if not 'Ranking' in dataframe.columns:
         dataframe.insert(0, 'Ranking', [i for i in range(1,len(dataframe)+1)])
@@ -226,33 +261,56 @@ def rank_dataframe(dataframe):
     return dataframe
 
 
+def run_update_dataset():
+    for i in range(0, len(rl_envs)):
+        rl_env = rl_envs[i]
+        update_leaderboard_dataset(rl_env["rl_env"])
+
+
+def filter_data(rl_env, path, user_id):
+    print("RL ENV", rl_env)
+    print("PATH", path)
+    print("USER ID", user_id)
+    data_df = get_data_no_html(rl_env, path)
+    print(data_df)
+    models = []
+    models = data_df[data_df["User"] == user_id]
+
+    for index, row in models.iterrows():
+        user_id = row["User"]
+        models.loc[index, "User"] = make_clickable_user(user_id)
+        model_id = row["Model"]
+        models.loc[index, "Model"] = make_clickable_model(model_id)
+        
+
+    print(models)
+    return models
+    
 with block:
     gr.Markdown(f"""
     # 🏆 The Deep Reinforcement Learning Course Leaderboard 🏆 
     
-    This is the leaderboard of trained agents during the Deep Reinforcement Learning Course. A free course from beginner to expert.
+    This is the leaderboard of trained agents during the <a href="https://huggingface.co/learn/deep-rl-course/unit0/introduction?fw=pt">Deep Reinforcement Learning Course</a>. A free course from beginner to expert.
     
-    Just choose which environment you trained your agent on and with Ctrl+F find your rank 🏆
-
-    **The leaderboard is updated every hour. If you don't find your model, go to the bottom of the page and click on the refresh button**
+    ### We only display the best 100 models
+    If you want to **find yours, type your user id and click on Search my models.**
+    You **can click on the model's name** to be redirected to its model card, including documentation.
     
+    ### How are the results calculated?
     We use **lower bound result to sort the models: mean_reward - std_reward.**
 
-    You **can click on the model's name** to be redirected to its model card which includes documentation.
+    ### I can't find my model 😭
+    The leaderboard is **updated every hour** if you can't find your models, just wait for the next update.
     
+    ### The Deep RL Course
     🤖 You want to try to train your agents? <a href="https://huggingface.co/deep-rl-course/unit0/introduction?fw=pt" target="_blank"> Check the Hugging Face free Deep Reinforcement Learning Course 🤗 </a>.
-    
-    You want to compare two agents? <a href="https://huggingface.co/spaces/ThomasSimonini/Compare-Reinforcement-Learning-Agents" target="_blank">It's possible using this Spaces demo 👀 </a>.
-    
+        
     🔧 There is an **environment missing?** Please open an issue.
-
-    For the RL course progress check out <a href="https://huggingface.co/spaces/ThomasSimonini/Check-my-progress-Deep-RL-Course" target="_blank"> User Progress App </a>
     """)
-    
-    #for rl_env in RL_ENVS:
+    path_ = download_leaderboard_dataset()
+
     for i in range(0, len(rl_envs)):
         rl_env = rl_envs[i]
-       
         with gr.TabItem(rl_env["rl_env_beautiful"]) as rl_tab:
             with gr.Row():
                 markdown = """
@@ -260,14 +318,25 @@ with block:
                     
                     """.format(name_leaderboard = rl_env["rl_env_beautiful"], video_link = rl_env["video_link"])
                 gr.Markdown(markdown)
+                #gr.Textbox(value=filter_data(rl_env["rl_env"], path_))
+            
             with gr.Row():
-                rl_env["global"] = gr.components.Dataframe(value= get_model_dataframe(rl_env["rl_env"]), headers=["Ranking 🏆", "User 🤗", "Model id 🤖", "Results", "Mean Reward", "Std Reward"], datatype=["number", "markdown", "markdown", "number", "number", "number"])
+                gr.Markdown("""
+                    ## Search your models
+                    Simply type your user id to find your models
+                    """)
+                
             with gr.Row():
-                data_run = gr.Button("Refresh")
-                #print("rl_env", rl_env["rl_env"])
-                val = gr.Variable(value=[rl_env["rl_env"]])
-                data_run.click(get_model_dataframe, inputs=[val], outputs =rl_env["global"])
-            
+                user_id = gr.Textbox(label= "Your user id")
+                search_btn = gr.Button("Search my models 🔎")
+                env = gr.Variable(rl_env["rl_env"])
+                grpath = gr.Variable(path_)
+            with gr.Row():
+                gr_dataframe = gr.components.Dataframe(value=get_data(rl_env["rl_env"], path_), headers=["Ranking 🏆", "User 🤗", "Model id 🤖", "Results", "Mean Reward", "Std Reward"], datatype=["number", "markdown", "markdown", "number", "number", "number"],  row_count=(100, 'fixed'))
+             
+            with gr.Row():
+                #gr_search_dataframe = gr.components.Dataframe(headers=["Ranking 🏆", "User 🤗", "Model id 🤖", "Results", "Mean Reward", "Std Reward"], datatype=["number", "markdown", "markdown", "number", "number", "number"], visible=False)
+                search_btn.click(fn=filter_data, inputs=[env, grpath, user_id], outputs=gr_dataframe, api_name="filter_data")
                        
 block.launch()
 
@@ -281,9 +350,10 @@ def refresh_leaderboard():
         temp = get_model_dataframe(rl_env)
         rl_env["global"] = temp
     print("The leaderboard has been updated")
+                       
+block.launch()
 
 scheduler = BackgroundScheduler()
 # Refresh every hour
-scheduler.add_job(func=refresh_leaderboard, trigger="interval", seconds=3600)
-scheduler.start()
-
+scheduler.add_job(func=run_update_dataset, trigger="interval", seconds=3600)
+scheduler.start()