edit colab (+1 squashed commits)

Squashed commits:

[c7ccb99d] update colab with llava

class.py

@@ -273,7 +273,7 @@ class model_backend(InferenceModel):
         unbantokens=False, bantokens=None, usemirostat=None, forceversion=0, nommap=self.kcpp_nommap,
         usemlock=False, noavx2=self.kcpp_noavx2, debugmode=self.kcpp_debugmode, skiplauncher=True, hordeconfig=None, noblas=self.kcpp_noblas,
         useclblast=self.kcpp_useclblast, usecublas=self.kcpp_usecublas, usevulkan=self.kcpp_usevulkan, gpulayers=self.kcpp_gpulayers, tensor_split=self.kcpp_tensor_split, config=None,
-        onready='', multiuser=False, foreground=False, preloadstory=None, noshift=False, remotetunnel=False, ssl=False, benchmark=False, nocertify=False, sdconfig=None, mmproj=None)
+        onready='', multiuser=False, foreground=False, preloadstory=None, noshift=False, remotetunnel=False, ssl=False, benchmark=None, nocertify=False, sdconfig=None, mmproj=None)
 
 
         #koboldcpp.main(kcppargs,False) #initialize library without enabling Lite http server

colab.ipynb

@@ -52,6 +52,11 @@
         "Layers = 99 #@param [99]{allow-input: true}\r\n",
         "ContextSize = 4096 #@param [4096] {allow-input: true}\r\n",
         "ForceRebuild = False #@param {type:\"boolean\"}\r\n",
+        "#@markdown <hr>\r\n",
+        "LoadLLaVAmmproj = False #@param {type:\"boolean\"}\r\n",
+        "LLaVAmmproj = \"https://huggingface.co/concedo/koboldcpp-mmproj/resolve/main/llama-13b-mmproj-v1.5.Q4_1.gguf\" #@param [\"https://huggingface.co/concedo/koboldcpp-mmproj/resolve/main/llama-13b-mmproj-v1.5.Q4_1.gguf\"]{allow-input: true}\r\n",
+        "Vcommand = \"\"\r\n",
+        "#@markdown <hr>\r\n",
         "LoadImgModel = False #@param {type:\"boolean\"}\r\n",
         "ImgModel = \"https://huggingface.co/koboldcpp/imgmodel/resolve/main/imgmodel_older_q4_0.gguf\" #@param [\"https://huggingface.co/koboldcpp/imgmodel/resolve/main/imgmodel_older_q4_0.gguf\"]{allow-input: true}\r\n",
         "SCommand = \"\"\r\n",
@@ -67,6 +72,10 @@
         "kvers = kvers[0]\r\n",
         "if ForceRebuild:\r\n",
         "  kvers = \"force_rebuild\"\r\n",
+        "if LLaVAmmproj and LoadLLaVAmmproj:\r\n",
+        "  Vcommand = \"--mmproj vmodel.gguf\"\r\n",
+        "else:\r\n",
+        "  SCommand = \"\"\r\n",
         "if ImgModel and LoadImgModel:\r\n",
         "  SCommand = \"--sdconfig imodel.gguf clamped 4 quant\"\r\n",
         "else:\r\n",
@@ -79,9 +88,11 @@
         "!apt update\r\n",
         "!apt install aria2 -y\r\n",
         "!aria2c -x 10 -o model.gguf --summary-interval=5 --download-result=default --allow-overwrite=true --file-allocation=none $Model\r\n",
+        "if VCommand:\r\n",
+        "  !aria2c -x 10 -o vmodel.gguf --summary-interval=5 --download-result=default --allow-overwrite=true --file-allocation=none $LLaVAmmproj\r\n",
         "if SCommand:\r\n",
         "  !aria2c -x 10 -o imodel.gguf --summary-interval=5 --download-result=default --allow-overwrite=true --file-allocation=none $ImgModel\r\n",
-        "!python koboldcpp.py model.gguf --usecublas 0 mmq --multiuser --gpulayers $Layers --contextsize $ContextSize --quiet --remotetunnel $SCommand\r\n"
+        "!python koboldcpp.py model.gguf --usecublas 0 mmq --multiuser --gpulayers $Layers --contextsize $ContextSize --quiet --remotetunnel $Vcommand $SCommand\r\n"
       ]
     }
   ],

koboldcpp.py

@@ -2870,7 +2870,7 @@ def main(launch_args,start_server=True):
         timer_thread = threading.Timer(1, onready_subprocess) #1 second delay
         timer_thread.start()
 
-    if args.model_param and args.benchmark is not None:
+    if args.model_param and args.benchmark:
         from datetime import datetime, timezone
         global libname
         start_server = False