From df7c2b9923665390d294175b9954757a3da838ad Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Mon, 11 Nov 2024 19:40:47 +0800
Subject: [PATCH] renamed some labels

---
 colab.ipynb    |  8 ++++----
 kcpp_docs.embd |  8 ++++++++
 klite.embd     | 35 ++++++++++++++++++++++++++++-------
 koboldcpp.py   |  6 +++---
 4 files changed, 43 insertions(+), 14 deletions(-)

diff --git a/colab.ipynb b/colab.ipynb
index efb844490..61ead6a00 100644
--- a/colab.ipynb
+++ b/colab.ipynb
@@ -54,8 +54,8 @@
         "FlashAttention = True #@param {type:\"boolean\"}\r\n",
         "FACommand = \"\"\r\n",
         "#@markdown <hr>\r\n",
-        "LoadLLaVAmmproj = False #@param {type:\"boolean\"}\r\n",
-        "LLaVAmmproj = \"https://huggingface.co/koboldcpp/mmproj/resolve/main/llama-13b-mmproj-v1.5.Q4_1.gguf\" #@param [\"https://huggingface.co/koboldcpp/mmproj/resolve/main/llama-13b-mmproj-v1.5.Q4_1.gguf\",\"https://huggingface.co/koboldcpp/mmproj/resolve/main/mistral-7b-mmproj-v1.5-Q4_1.gguf\",\"https://huggingface.co/koboldcpp/mmproj/resolve/main/llama-7b-mmproj-v1.5-Q4_0.gguf\",\"https://huggingface.co/koboldcpp/mmproj/resolve/main/LLaMA3-8B_mmproj-Q4_1.gguf\"]{allow-input: true}\r\n",
+        "LoadVisionMMProjector = False #@param {type:\"boolean\"}\r\n",
+        "Mmproj = \"https://huggingface.co/koboldcpp/mmproj/resolve/main/llama-13b-mmproj-v1.5.Q4_1.gguf\" #@param [\"https://huggingface.co/koboldcpp/mmproj/resolve/main/llama-13b-mmproj-v1.5.Q4_1.gguf\",\"https://huggingface.co/koboldcpp/mmproj/resolve/main/mistral-7b-mmproj-v1.5-Q4_1.gguf\",\"https://huggingface.co/koboldcpp/mmproj/resolve/main/llama-7b-mmproj-v1.5-Q4_0.gguf\",\"https://huggingface.co/koboldcpp/mmproj/resolve/main/LLaMA3-8B_mmproj-Q4_1.gguf\"]{allow-input: true}\r\n",
         "VCommand = \"\"\r\n",
         "#@markdown <hr>\r\n",
         "LoadImgModel = False #@param {type:\"boolean\"}\r\n",
@@ -71,7 +71,7 @@
         "  raise RuntimeError(\"⚠️Colab did not give you a GPU due to usage limits, this can take a few hours before they let you back in. Check out https://lite.koboldai.net for a free alternative (that does not provide an API link but can load KoboldAI saves and chat cards) or subscribe to Colab Pro for immediate access.⚠️\")\r\n",
         "\r\n",
         "%cd /content\r\n",
-        "if LLaVAmmproj and LoadLLaVAmmproj:\r\n",
+        "if Mmproj and LoadVisionMMProjector:\r\n",
         "  VCommand = \"--mmproj vmodel.gguf\"\r\n",
         "else:\r\n",
         "  SCommand = \"\"\r\n",
@@ -99,7 +99,7 @@
         "    Model = Model.replace(\"/blob/main/\", \"/resolve/main/\")\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",
+        "  !aria2c -x 10 -o vmodel.gguf --summary-interval=5 --download-result=default --allow-overwrite=true --file-allocation=none $Mmproj\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",
         "if WCommand:\r\n",
diff --git a/kcpp_docs.embd b/kcpp_docs.embd
index 647d2fad7..a32ed0433 100644
--- a/kcpp_docs.embd
+++ b/kcpp_docs.embd
@@ -1272,6 +1272,7 @@
                                   "width": 512,
                                   "height": 512,
                                   "seed": -1,
+                                  "clip_skip": -1,
                                   "sampler_name": "Euler a"
                                },
                                "schema": {
@@ -1297,6 +1298,9 @@
                                      "seed": {
                                         "type": "number"
                                      },
+                                     "clip_skip": {
+                                        "type": "number"
+                                     },
                                      "sampler_name": {
                                         "type": "string"
                                      },
@@ -1356,6 +1360,7 @@
                                   "width": 512,
                                   "height": 512,
                                   "seed": -1,
+                                  "clip_skip": -1,
                                   "sampler_name": "Euler a",
                                   "denoising_strength": 0.6,
                                   "init_images":["base64_image_data"],
@@ -1383,6 +1388,9 @@
                                      "seed": {
                                         "type": "number"
                                      },
+                                     "clip_skip": {
+                                        "type": "number"
+                                     },
                                      "sampler_name": {
                                         "type": "string"
                                      },
diff --git a/klite.embd b/klite.embd
index 8ff5ebce0..e12982e9f 100644
--- a/klite.embd
+++ b/klite.embd
@@ -12,7 +12,7 @@ Current version indicated by LITEVER below.
 -->
 
 <script>
-	const LITEVER = 185;
+	const LITEVER = 186;
 	const urlParams = new URLSearchParams(window.location.search);
 	var localflag = true;
 	const STORAGE_PREFIX = (localflag?"e_":"")+"kaihordewebui_";
@@ -4355,6 +4355,7 @@ Current version indicated by LITEVER below.
 	const XTTS_ID = 1000;
 	const ALLTALK_ID = 1001;
 	const OAI_TTS_ID = 1002;
+	const BIG_HD_RES_PX = 512; //when saving oversized images, could be 640?
 	const HD_RES_PX = 512;
 	const NO_HD_RES_PX = 320;
 	const SAVE_SLOTS = 6;
@@ -4527,6 +4528,7 @@ Current version indicated by LITEVER below.
 		img_allowhd: true,
 		img_crop: false,
 		img_img2imgstr: 0.6,
+		img_clipskip: -1,
 		img_steps: 20,
 		img_sampler: "Euler a",
 		img_aspect:0, //0=square,1=portrait,2=landscape,3=bigsquare
@@ -5570,6 +5572,10 @@ Current version indicated by LITEVER below.
 			a1111_t2i_payload.init_images = [req_payload.source_image];
 			a1111_t2i_payload.denoising_strength = req_payload.params.denoising_strength;
 		}
+		if(req_payload.params.clip_skip && req_payload.params.clip_skip>0)
+		{
+			a1111_t2i_payload.clip_skip = req_payload.params.clip_skip;
+		}
 
 		if(localsettings.save_remote_images)
 		{
@@ -10485,6 +10491,7 @@ Current version indicated by LITEVER below.
 		document.getElementById("save_remote_images").checked = localsettings.save_remote_images;
 		document.getElementById("img_cfgscale").value = localsettings.img_cfgscale;
 		document.getElementById("img_img2imgstr").value = localsettings.img_img2imgstr;
+		document.getElementById("img_clipskip").value = localsettings.img_clipskip;
 		document.getElementById("img_aspect").value = localsettings.img_aspect;
 		document.getElementById("img_sampler").value = localsettings.img_sampler;
 		document.getElementById("img_steps").value = localsettings.img_steps;
@@ -10785,6 +10792,7 @@ Current version indicated by LITEVER below.
 
 		localsettings.img_cfgscale = parseFloat(document.getElementById("img_cfgscale").value);
 		localsettings.img_img2imgstr = parseFloat(document.getElementById("img_img2imgstr").value);
+		localsettings.img_clipskip = parseInt(document.getElementById("img_clipskip").value);
 		localsettings.img_aspect = parseInt(document.getElementById("img_aspect").value);
 		localsettings.img_sampler = document.getElementById("img_sampler").value;
 		localsettings.img_steps = parseInt(document.getElementById("img_steps").value);
@@ -10800,7 +10808,12 @@ Current version indicated by LITEVER below.
 		{
 			localsettings.img_img2imgstr = defaultsettings.img_img2imgstr;
 		}
+		if(isNaN(localsettings.img_clipskip))
+		{
+			localsettings.img_clipskip = defaultsettings.img_clipskip;
+		}
 		localsettings.img_img2imgstr = cleannum(localsettings.img_img2imgstr, 0.0, 1.0);
+		localsettings.img_clipskip = cleannum(localsettings.img_clipskip, -1, 20);
 		if(isNaN(localsettings.img_aspect))
 		{
 			localsettings.img_aspect = defaultsettings.img_aspect;
@@ -13881,6 +13894,10 @@ Current version indicated by LITEVER below.
 			genimg_payload["source_image"] = base64img;
 			genimg_payload["params"]["denoising_strength"] = localsettings.img_img2imgstr;
 		}
+		if(localsettings.img_clipskip>0)
+		{
+			genimg_payload["params"]["clip_skip"] = localsettings.img_clipskip;
+		}
 
 		if(localsettings.generate_images_mode==1) //horde
 		{
@@ -13931,7 +13948,7 @@ Current version indicated by LITEVER below.
 				{
 					//console.log(outputimg);
 					let origImg = "data:image/jpeg;base64," + outputimg;
-					let imgres = localsettings.img_allowhd?HD_RES_PX:NO_HD_RES_PX;
+					let imgres = localsettings.img_allowhd?(localsettings.img_aspect==0?HD_RES_PX:BIG_HD_RES_PX):NO_HD_RES_PX;
 					compressImage(origImg, (newDataUri) => {
 						image_db[imgid].done = true;
 						image_db[imgid].result = newDataUri;
@@ -13961,7 +13978,7 @@ Current version indicated by LITEVER below.
 					{
 						//console.log(outputimg);
 						let origImg = "data:image/jpeg;base64," + outputimg;
-						let imgres = localsettings.img_allowhd?HD_RES_PX:NO_HD_RES_PX;
+						let imgres = localsettings.img_allowhd?(localsettings.img_aspect==0?HD_RES_PX:BIG_HD_RES_PX):NO_HD_RES_PX;
 						compressImage(origImg, (newDataUri) => {
 							image_db[imgid].done = true;
 							image_db[imgid].result = newDataUri;
@@ -14610,7 +14627,7 @@ Current version indicated by LITEVER below.
 		console.log("polling for pending images " + imagecount);
 		for (let key in image_db) {
 			let img = image_db[key];
-			if (img.done == false && img.poll_category==1) {
+			if (img.done == false && img.poll_category==1) { //horde image polling
 				//call check
 				fetch(stablehorde_poll_endpoint + "/" + key)
 					.then(x => x.json())
@@ -14637,7 +14654,7 @@ Current version indicated by LITEVER below.
 										img.queue = 0;
 										let origImg = "data:image/jpeg;base64," + finalimg.generations[0].img;
 										//console.log("Original image: " + origImg);
-										let imgres = localsettings.img_allowhd?HD_RES_PX:NO_HD_RES_PX;
+										let imgres = localsettings.img_allowhd?(localsettings.img_aspect==0?HD_RES_PX:BIG_HD_RES_PX):NO_HD_RES_PX;
 										compressImage(origImg, (newDataUri) => {
 											img.result = newDataUri;
 										}, true, false, imgres,0.35,false);
@@ -14660,7 +14677,7 @@ Current version indicated by LITEVER below.
 						delete image_db[key];
 					});
 			}
-			else if (img.done == false && img.poll_category==2)
+			else if (img.done == false && img.poll_category==2) //comfyui image polling
 			{
 				//comfyui polling
 				fetch(localsettings.saved_comfy_url + comfy_history_endpoint + "/" + key, {
@@ -14687,7 +14704,7 @@ Current version indicated by LITEVER below.
 							reader.onloadend = () => {
 								img.queue = 0;
 								let origImg = reader.result;
-								let imgres = localsettings.img_allowhd?HD_RES_PX:NO_HD_RES_PX;
+								let imgres = localsettings.img_allowhd?(localsettings.img_aspect==0?HD_RES_PX:BIG_HD_RES_PX):NO_HD_RES_PX;
 								compressImage(origImg, (newDataUri) => {
 									img.result = newDataUri;
 								}, true, false, imgres,0.35,false);
@@ -19571,6 +19588,10 @@ Current version indicated by LITEVER below.
 								</span>: </div>
 								<input title="Img2Img Strength" type="text" inputmode="decimal" id="img_img2imgstr" style="width:60px">
 							</div>
+							<div class="inlinelabel">
+								<div class="justifyleft" style="padding:4px">Clip Skip: </div>
+								<input title="Clip Skip" type="text" inputmode="decimal" id="img_clipskip" style="width:60px">
+							</div>
 							<div class="inlinelabel">
 								<div class="justifyleft" style="padding:4px">Save Higher-Res <span class="helpicon">?
 									<span class="helptext">This option will result in larger save files which may be slower. Changing this setting only applies to NEW images.</span>
diff --git a/koboldcpp.py b/koboldcpp.py
index 04ff99354..ca9992d58 100644
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -2102,7 +2102,7 @@ Enter Prompt:<br>
                     self.send_response(503)
                     self.end_headers(content_type='application/json')
                     self.wfile.write(json.dumps({"detail": {
-                            "msg": "No LLaVA model loaded",
+                            "msg": "No Vision model loaded",
                             "type": "service_unavailable",
                         }}).encode())
                     return
@@ -2982,7 +2982,7 @@ def show_gui():
     makefileentry(model_tab, "Model:", "Select GGUF or GGML Model File", model_var, 1,width=280, onchoosefile=on_picked_model_file,tooltiptxt="Select a GGUF or GGML model file on disk to be loaded.")
     makefileentry(model_tab, "Lora:", "Select Lora File",lora_var, 3,width=280,tooltiptxt="Select an optional GGML LoRA adapter to use.\nLeave blank to skip.")
     makefileentry(model_tab, "Lora Base:", "Select Lora Base File", lora_base_var, 5,width=280,tooltiptxt="Select an optional F16 GGML LoRA base file to use.\nLeave blank to skip.")
-    makefileentry(model_tab, "LLaVA mmproj:", "Select LLaVA mmproj File", mmproj_var, 7,width=280,tooltiptxt="Select a mmproj file to use for LLaVA.\nLeave blank to skip.")
+    makefileentry(model_tab, "Vision mmproj:", "Select Vision mmproj File", mmproj_var, 7,width=280,tooltiptxt="Select a mmproj file to use for vision models like LLaVA.\nLeave blank to skip.")
     makefileentry(model_tab, "Preloaded Story:", "Select Preloaded Story File", preloadstory_var, 9,width=280,tooltiptxt="Select an optional KoboldAI JSON savefile \nto be served on launch to any client.")
     makefileentry(model_tab, "ChatCompletions Adapter:", "Select ChatCompletions Adapter File", chatcompletionsadapter_var, 12, width=250, filetypes=[("JSON Adapter", "*.json")], tooltiptxt="Select an optional ChatCompletions Adapter JSON file to force custom instruct tags.")
     def pickpremadetemplate():
@@ -4661,7 +4661,7 @@ if __name__ == '__main__':
     advparser.add_argument("--quiet", help="Enable quiet mode, which hides generation inputs and outputs in the terminal. Quiet mode is automatically enabled when running a horde worker.", action='store_true')
     advparser.add_argument("--ssl", help="Allows all content to be served over SSL instead. A valid UNENCRYPTED SSL cert and key .pem files must be provided", metavar=('[cert_pem]', '[key_pem]'), nargs='+')
     advparser.add_argument("--nocertify", help="Allows insecure SSL connections. Use this if you have cert errors and need to bypass certificate restrictions.", action='store_true')
-    advparser.add_argument("--mmproj", help="Select a multimodal projector file for LLaVA.", default="")
+    advparser.add_argument("--mmproj", help="Select a multimodal projector file for vision models like LLaVA.", default="")
     advparser.add_argument("--password", help="Enter a password required to use this instance. This key will be required for all text endpoints. Image endpoints are not secured.", default=None)
     advparser.add_argument("--ignoremissing", help="Ignores all missing non-essential files, just skipping them instead.", action='store_true')
     advparser.add_argument("--chatcompletionsadapter", help="Select an optional ChatCompletions Adapter JSON file to force custom instruct tags.", default="")