adjust max ctx back to 12288, allow img2img with ref images

koboldcpp.py

@@ -61,13 +61,14 @@ default_vae_tile_threshold = 640
 default_sdvaedevice = 'main'
 default_sdclipdevice = 'CPU'
 default_native_ctx = 16384
-default_genlen = 2048
+default_genlen = 1536
 overridekv_max = 16
 default_autofit_padding = 1024
 lora_filenames_max = 4
 multiuser_concurrent_limit = 10
 swa_padding_default = 0
 default_reqtimeout = 600 # 10 min default
+default_maxctx = 12288
 
 # abuse prevention
 stop_token_max = 256
@@ -77,7 +78,7 @@ dry_seq_break_max = 128
 extra_images_max = 4 # for kontext/qwen img
 
 # global vars
-KcppVersion = "1.115.1"
+KcppVersion = "1.115.2"
 showdebug = True
 kcpp_instance = None #global running instance
 global_memory = {"tunnel_url": "", "restart_target":"", "input_to_exit":False, "load_complete":False, "restart_override_base_config":"", "last_active_timestamp":datetime.now(), "triggered_sleeping":False, "current_model":"initial_model", "base_config":"", "swapReqType": None, "autoswapmode": False}
@@ -109,7 +110,7 @@ imglora_bypath = {}    # len(imglora_bypath) == 0 <==> static loras
 imglora_name2path = {}
 imglora_cached = True
 imglora_initial_fixed = True
-maxctx = 16384
+maxctx = default_maxctx
 maxhordectx = 0 #set to whatever maxctx is if 0
 maxhordelen = 1024
 modelbusy = threading.Lock()
@@ -8568,7 +8569,7 @@ def show_gui():
         makecheckbox(quick_tab, name, properties[0], int(idx/2) + 20, idx % 2, tooltiptxt=properties[1])
 
     # context size
-    makeslider(quick_tab, "Context Size:", contextsize_text, context_var, 40, width=280, set=17, 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.")
+    makeslider(quick_tab, "Context Size:", contextsize_text, context_var, 40, width=280, set=13, 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.")
 
     # load model
     makefileentry(quick_tab, "GGUF Text Model:", "Select GGUF or GGML Model File", model_var, 50, 280, onchoosefile=on_picked_model_file,tooltiptxt="Select a GGUF or GGML model file on disk to be loaded.")
@@ -8645,7 +8646,7 @@ def show_gui():
     makelabelentry(context_tab, "CacheSlots:", smartcacheslots_var, row=5, padx=(300), singleline=True, tooltip="Number of slots for smartcache",labelpadx=(220))
 
     # context size
-    makeslider(context_tab, "Context Size:",contextsize_text, context_var, 18, width=280, set=17,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.")
+    makeslider(context_tab, "Context Size:",contextsize_text, context_var, 18, width=280, set=13,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.")
     context_var.trace_add("write", changed_gpulayers_estimate)
     makelabelentry(context_tab, "Default Gen Amt:", defaultgenamt_var, row=20, padx=(120), singleline=True, tooltip="How many tokens to generate by default, if not specified. Must be smaller than context size. Usually, your frontend GUI will override this.")
     makelabelentry(context_tab, "Prompt Limit:", genlimit_var, row=20, padx=(300), singleline=True, tooltip="If set, restricts max output tokens to this limit regardless of API request. Set to 0 to disable.",labelpadx=(210))
@@ -12020,7 +12021,7 @@ if __name__ == '__main__':
     modelgroup.add_argument("--model","-m", metavar=('[filenames]'), help="Model file to load. Accepts multiple values if they are URLs.", type=str, nargs='+', default=[])
     modelgroup.add_argument("model_param", help="Model file to load (positional)", nargs="?")
     parser.add_argument("--config", metavar=('[filename]'), help="Load settings from a .kcpps file. Other arguments will be ignored", type=str, nargs=1)
-    parser.add_argument("--contextsize","--ctx-size", "-c", help="Controls the memory allocated for maximum context size, only change if you need more RAM for big contexts. (default 16384).",metavar=('[256 to 262144]'), type=check_range(int,256,262144), default=16384)
+    parser.add_argument("--contextsize","--ctx-size", "-c", help=f"Controls the memory allocated for maximum context size, only change if you need more RAM for big contexts. (default {default_maxctx}).",metavar=('[256 to 262144]'), type=check_range(int,256,262144), default=default_maxctx)
     parser.add_argument("--gpulayers","--gpu-layers","--n-gpu-layers","-ngl", help="Set number of layers to offload to GPU (when using GPU). Set to -1 to enable autofit (default), set to 0 to disable GPU offload.",metavar=('[GPU layers]'), nargs='?', const=1, type=int, default=-1)
     parser.add_argument("--host", metavar=('[ipaddr]'), help="Host IP to listen on. If this flag is not set, all routable interfaces are accepted.", default="")
     parser.add_argument("--launch", help="Launches a web browser when load is completed.", action='store_true')