You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: koboldcpp.py
+24-20
Original file line number
Diff line number
Diff line change
@@ -2511,15 +2511,18 @@ def changerunmode(a,b,c):
2511
2511
quick_mmq_box=makecheckbox(quick_tab, "Use QuantMatMul (mmq)", mmq_var, 4,1,tooltiptxt="Enable MMQ mode instead of CuBLAS for prompt processing. Read the wiki. Speed may vary.")
"Launch Browser": {"variable": launchbrowser, "description": "Launches your default browser after model loading is complete"},
2516
+
"Disable MMAP": {"variable": disablemmap, "description": "Avoids using mmap to load models if enabled"},
2517
+
"Use ContextShift": {"variable": contextshift, "description": "Uses Context Shifting to reduce reprocessing.\nRecommended. Check the wiki for more info."},
2518
+
"Remote Tunnel": {"variable": remotetunnel, "description": "Creates a trycloudflare tunnel.\nAllows you to access koboldcpp from other devices over an internet URL."},
makeslider(quick_tab, "Context Size:", contextsize_text, context_var, 0, len(contextsize_text)-1, 30, width=280, set=5,tooltip="What is the maximum context size to support. Model specific. You cannot exceed it.\nLarger contexts require more memory, and not all models support it.")
2525
2528
@@ -2555,17 +2558,18 @@ def changerunmode(a,b,c):
2555
2558
makelabelentry(hardware_tab, "Threads:" , threads_var, 11, 50,tooltip="How many threads to use.\nRecommended value is your CPU core count, defaults are usually OK.")
"Launch Browser": {"variable": launchbrowser, "description": "Launches your default browser after model loading is complete"},
2563
+
"High Priority": {"variable": highpriority, "description": "Increases the koboldcpp process priority.\nMay cause lag or slowdown instead. Not recommended."},
2564
+
"Disable MMAP": {"variable": disablemmap, "description": "Avoids using mmap to load models if enabled"},
2565
+
"Use mlock": {"variable": usemlock, "description": "Enables mlock, preventing the RAM used to load the model from being paged out."},
2566
+
"Debug Mode": {"variable": debugmode, "description": "Enables debug mode, with extra info printed to the terminal."},
2567
+
"Keep Foreground": {"variable": keepforeground, "description": "Bring KoboldCpp to the foreground every time there is a new generation."}
makelabelentry(hardware_tab, "BLAS threads:" , blas_threads_var, 14, 50,tooltip="How many threads to use during BLAS processing.\nIf left blank, uses same value as regular thread count.")
0 commit comments