Add vizdoom

app.py

@@ -39,6 +39,23 @@ def get_user_models(hf_username, env_tag, lib_tag):
     return user_model_ids
 
 
+def get_user_sf_models(hf_username, env_tag, lib_tag):
+    models_sf = []
+    models = api.list_models(author=hf_username, filter=["reinforcement-learning", lib_tag])
+
+    user_model_ids = [x.modelId for x in models]
+
+    for model in user_model_ids:
+        meta = get_metadata(model)
+        if meta is None:
+            continue
+        result = meta["model-index"][0]["results"][0]["dataset"]["name"]
+        if result == env_tag:
+            models_sf.append(model)
+            
+    return models_sf
+
+
 def get_metadata(model_id):
   """
   Get model metadata (contains evaluation data)
@@ -225,18 +242,22 @@ def certification(hf_username, first_name, last_name):
     },
       {
           "unit": "Unit 8 PII",
-          "env": "Vizdoom-Battle",
-          "library": "cleanrl",
-          "min_result": 100,
+          "env": "doom_health_gathering_supreme",
+          "library": "sample-factory",
+          "min_result": 5,
           "best_result": 0,
           "best_model_id": "",
           "passed_": False
     },
   ] 
   for unit in results_certification:
+    if unit["unit"] != "Unit 8 PII":
     # Get user model
-    user_models = get_user_models(hf_username, unit['env'], unit['library'])
-
+        user_models = get_user_models(hf_username, unit['env'], unit['library'])
+    # For sample factory vizdoom we don't have env tag for now
+    else: 
+      user_models = get_user_sf_models(hf_username, unit['env'], unit['library'])
+    
     # Calculate the best result and get the best_model_id
     best_result, best_model_id = calculate_best_result(user_models)