aliveness calculation fixes

app_config.py

@@ -20,11 +20,11 @@ ENDPOINT_NAMES = {
     # "CTL_llama2": "texter_simulator",
     "CTL_llama3": {
         "name": "texter_simulator_llm",
-        "model_type": "llm"
+        "model_type": "text-generation"
     },
     # "CTL_llama3": {
     #     "name": "databricks-meta-llama-3-1-70b-instruct",
-    #     "model_type": "llm"
+    #     "model_type": "text-generation"
     # },
     # 'CTL_llama2': "llama2_convo_sim",
     # "CTL_mistral": "convo_sim_mistral",
@@ -38,7 +38,7 @@ ENDPOINT_NAMES = {
     },
     "training_adherence": {
         "name": "training_adherence",
-        "model_type": "llm"
+        "model_type": "text-completion"
     },
 }
 

pages/model_loader.py

@@ -18,16 +18,16 @@ MODELS2LOAD = {
     "training_adherence": {"model_name": "Training Adherence", "loaded":None},
 }
 
-def write_model_status(writer, model_name, loaded, fail=None):
-    if loaded:
+def write_model_status(writer, model_name, loaded, fail=False):
+    if loaded == "200":
         writer.write(f"✅ - {model_name} Loaded")
-    else:
-        if fail in ["400", "500"]:
+    if fail:
+        if loaded in ["400", "500"]:
             writer.write(f"❌ - {model_name} Failed to Load, Contact ifbarrerarincon@crisistextline.org")
-        elif fail == "404":
+        elif loaded == "404":
             writer.write(f"❌ - {model_name} Still loading, please try in a couple of minutes")
-        else:
-            writer.write(f"🔄 - {model_name} Loading")
+    else:
+        writer.write(f"🔄 - {model_name} Loading")
 
 with st.status("Loading Models Please Wait...(this may take up to 5 min)", expanded=True) as status:
 
@@ -37,9 +37,9 @@ with st.status("Loading Models Please Wait...(this may take up to 5 min)", expan
     while not models_alive:
         time.sleep(2)
         for name, config in MODELS2LOAD.items():
-            config["loaded"] = is_model_alive(ENDPOINT_NAMES[name])
+            config["loaded"] = is_model_alive(**ENDPOINT_NAMES[name])
 
-        models_alive = all([x['loaded'] for x in MODELS2LOAD.values()])
+        models_alive = all([x['loaded']=="200" for x in MODELS2LOAD.values()])
         
         for _, config in MODELS2LOAD.items():
             write_model_status(**config)

utils/app_utils.py

@@ -72,12 +72,18 @@ def is_model_alive(name, timeout=2, model_type="classificator"):
             body_request = {
                 "inputs": [""]
             }
-        elif model_type == "llm":
+        elif model_type == "text-completion":
             body_request = {
                 "prompt": "",
                 "temperature": 0,
                 "max_tokens": 1,
             }
+        elif model_type == "text-generation":
+            body_request = {
+                "messages": [{"role":"user","content":""}],
+                "max_tokens": 1,
+                "temperature": 0
+            }
 
         else:
             raise Exception(f"Model Type {model_type} not supported")