t

app.py

@@ -97,6 +97,7 @@ def f(context, temperature, top_p, max_length, model_idx, SPACE_VERIFICATION_KEY
             if main_gpt_j_api_up:
                 # for this api, a length of > 250 instantly errors, so use a while loop or something
                 # that would fetch results in chunks of 250
+                # NOTE change so it uses previous generated input every time
                 generated_text = ""
                 while (max_length > 0):
                     payload = {"inputs": context, "parameters": {"max_new_tokens": 250, "temperature": temperature, "top_p": top_p}}
@@ -104,10 +105,10 @@ def f(context, temperature, top_p, max_length, model_idx, SPACE_VERIFICATION_KEY
                     context = json.loads(response.content.decode("utf-8"))#[0]['generated_text']
                     context = get_generated_text(context)
                     # handle inconsistent inference API
-                    if 'generated_text' in context[0]:
-                        context = context[0]['generated_text']
-                    else:
-                        context = context[0][0]['generated_text']
+                    # if 'generated_text' in context[0]:
+                    #     context = context[0]['generated_text']
+                    # else:
+                    #     context = context[0][0]['generated_text']
                         
                     generated_text += context
                     max_length -= 250