From 752e19a2bb261fe8de365a9455f942bdd29d845b Mon Sep 17 00:00:00 2001
From: Gnome Ann <>
Date: Wed, 10 Nov 2021 17:38:30 -0500
Subject: [PATCH 1/3] Fix vars.modeldim not always being set

---
 aiserver.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/aiserver.py b/aiserver.py
index 176d5acb..afa2d4be 100644
--- a/aiserver.py
+++ b/aiserver.py
@@ -637,12 +637,15 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly"]):
                     generator = model.generate
                 elif(vars.breakmodel):  # Use both RAM and VRAM (breakmodel)
                     model = AutoModelForCausalLM.from_pretrained(vars.model)
+                    vars.modeldim = int(model.transformer.hidden_size)
                     device_config(model)
                 else:
                     model = AutoModelForCausalLM.from_pretrained(vars.model)
+                    vars.modeldim = int(model.transformer.hidden_size)
                     generator = model.generate
             else:
                 model = AutoModelForCausalLM.from_pretrained(vars.model)
+                vars.modeldim = int(model.transformer.hidden_size)
                 generator = model.generate
         
         # Suppress Author's Note by flagging square brackets (Old implementation)

From 11b0291bc4e9cac0e2fa2956bede010016d41704 Mon Sep 17 00:00:00 2001
From: Gnome Ann <>
Date: Wed, 10 Nov 2021 17:47:14 -0500
Subject: [PATCH 2/3] Use model.transformer.embed_dim if
 model.transformer.hidden_size doesn't exist

---
 aiserver.py | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/aiserver.py b/aiserver.py
index afa2d4be..0f908021 100644
--- a/aiserver.py
+++ b/aiserver.py
@@ -591,6 +591,12 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly"]):
             return stopping_criteria
         transformers.generation_utils.GenerationMixin._get_stopping_criteria = new_get_stopping_criteria
 
+        def get_hidden_size_from_model(model):
+            try:
+                return int(model.transformer.hidden_size)
+            except:
+                return int(model.transformer.embed_dim)
+
         # If custom GPT Neo model was chosen
         if(vars.model == "NeoCustom"):
             model_config = open(vars.custmodpth + "/config.json", "r")
@@ -632,20 +638,20 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly"]):
             if(vars.hascuda):
                 if(vars.usegpu):
                     model = AutoModelForCausalLM.from_pretrained(vars.model, device=0)
-                    vars.modeldim = int(model.transformer.hidden_size)
+                    vars.modeldim = get_hidden_size_from_model(model)
                     model = model.to(0)
                     generator = model.generate
                 elif(vars.breakmodel):  # Use both RAM and VRAM (breakmodel)
                     model = AutoModelForCausalLM.from_pretrained(vars.model)
-                    vars.modeldim = int(model.transformer.hidden_size)
+                    vars.modeldim = get_hidden_size_from_model(model)
                     device_config(model)
                 else:
                     model = AutoModelForCausalLM.from_pretrained(vars.model)
-                    vars.modeldim = int(model.transformer.hidden_size)
+                    vars.modeldim = get_hidden_size_from_model(model)
                     generator = model.generate
             else:
                 model = AutoModelForCausalLM.from_pretrained(vars.model)
-                vars.modeldim = int(model.transformer.hidden_size)
+                vars.modeldim = get_hidden_size_from_model(model)
                 generator = model.generate
         
         # Suppress Author's Note by flagging square brackets (Old implementation)

From 1fadcbe1e3e0cd4f7b57a3151b980f0f0f7b5357 Mon Sep 17 00:00:00 2001
From: Gnome Ann <>
Date: Thu, 11 Nov 2021 00:18:46 -0500
Subject: [PATCH 3/3] Send allowsp command on connect instead of on startup

---
 aiserver.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/aiserver.py b/aiserver.py
index 0f908021..cd49b249 100644
--- a/aiserver.py
+++ b/aiserver.py
@@ -727,6 +727,8 @@ def do_connect():
     emit('from_server', {'cmd': 'connected', 'smandelete': vars.smandelete, 'smanrename': vars.smanrename})
     if(vars.remote):
         emit('from_server', {'cmd': 'runs_remotely'})
+    if(vars.allowsp):
+        emit('from_server', {'cmd': 'allowsp', 'data': vars.allowsp})
     
     if(not vars.gamestarted):
         setStartState()
@@ -970,8 +972,6 @@ def get_message(msg):
 #==================================================================#
 def setStartState():
     txt = "<span>Welcome to <span class=\"color_cyan\">KoboldAI</span>! You are running <span class=\"color_green\">"+getmodelname()+"</span>.<br/>"
-    if(vars.allowsp):
-        emit('from_server', {'cmd': 'allowsp', 'data': vars.allowsp}, broadcast=True)
     if(not vars.noai):
         txt = txt + "Please load a game or enter a prompt below to begin!</span>"
     else: