session state

app.py

@@ -86,31 +86,6 @@ CACHE_SIZE = 10000
 def params(config):
     pass
 
-    
-def build_inference_api():
-    """HF inference api"""
-    API_URL = "https://api-inference.huggingface.co/models/debatelab/argument-analyst"
-    headers = {} # {"Authorization": f"Bearer {st.secrets['api_token']}"}
-
-    def query(inputs: str, parameters):
-        payload = {
-            "inputs": inputs,
-            "parameters": parameters,
-            "options": {"wait_for_model": True},
-        }
-        data = json.dumps(payload)
-        response = requests.request("POST", API_URL, headers=headers, data=data)
-        content = response.content.decode("utf-8")
-        try:
-            # as json 
-            result_json = json.loads(content) 
-        except Exception:
-            result_json = {"error": content}
-
-        return result_json
-
-    return query
-
 
 @st.cache(allow_output_mutation=True)
 def aaac_fields():
@@ -271,10 +246,9 @@ def run_model(mode_set, user_input):
     :returns: output dict
     """
 
-
-    #inference = build_inference_api()
-    with st.spinner('Initializing pipeline'):
-        inference = pipeline(task="text2text-generation", model=MODEL)
+    if "inference" not in st.session_state:
+        with st.spinner('Initializing pipeline'):
+            st.session_state.inference = pipeline(task="text2text-generation", model=MODEL)
 
     current_input = user_input.copy()
     output = []
@@ -289,26 +263,7 @@ def run_model(mode_set, user_input):
                 inquire_prompt = inquire_prompt + (f"{to_key}: {from_key}: {current_input[from_key]}")
             # inquire model
             inputs = inquire_prompt
-            attempts = 0
-            out = None
-            while not out and attempts<MAX_API_CALLS:
-                attempts += 1
-                try:
-                    # api call
-                    out = inference(inputs, **INFERENCE_PARAMS)
-                    if not isinstance(out, list):
-                        raise ValueError('Response is not a list.')
-                except Exception:
-                    if attempts < MAX_API_CALLS:
-                        st.warning(
-                            f"HF Inference API call (attempt {attempts} of {MAX_API_CALLS}) has failed. Response: {out}. Trying again..."
-                        )
-                        out = None
-                    else:
-                        st.warning(
-                            f"HF Inference API call (attempt {attempts} of {MAX_API_CALLS}) has failed. Response: {out}. Stopping."
-                        )
-                        return None
+            out = st.session_state.inference(inputs, **INFERENCE_PARAMS)
             out = out[0]['generated_text']
             # cleanup formalization
             if to_key in ['premises_formalized','conclusion_formalized']:
@@ -327,9 +282,7 @@ def run_model(mode_set, user_input):
             current_input[to_key] = out
 
     return output
-
-
-            
+      
 
 def main():