embeddings model colab

Author: LostRuins

colab.ipynb

@@ -73,6 +73,10 @@
     "WavTokModel = \"https://huggingface.co/koboldcpp/tts/resolve/main/WavTokenizer-Large-75-Q4_0.gguf\" #@param [\"https://huggingface.co/koboldcpp/tts/resolve/main/WavTokenizer-Large-75-Q4_0.gguf\"]{allow-input: true}\n",
     "TTSCommand = \"\"\n",
     "#@markdown <hr>\n",
+    "LoadEmbeddingsModel = False #@param {type:\"boolean\"}\n",
+    "EmbeddingsModel = \"https://huggingface.co/yixuan-chia/snowflake-arctic-embed-s-GGUF/resolve/main/snowflake-arctic-embed-s-Q4_0.gguf\" #@param [\"https://huggingface.co/yixuan-chia/snowflake-arctic-embed-s-GGUF/resolve/main/snowflake-arctic-embed-s-Q4_0.gguf\"]{allow-input: true}\n",
+    "ECommand = \"\"\n",
+    "#@markdown <hr>\n",
     "#@markdown This enables saving stories directly to your google drive. You will have to grant permissions, and then you can access the saves from the \"KoboldCpp Server Storage\" option.\n",
     "AllowSaveToGoogleDrive = False #@param {type:\"boolean\"}\n",
     "SavGdriveCommand = \"\"\n",
@@ -124,6 +128,10 @@
     "  TTSCommand = \"--ttsmodel ttsmodel.bin --ttswavtokenizer ttswavtok.bin --ttsgpu\"\n",
     "else:\n",
     "  TTSCommand = \"\"\n",
+    "if EmbeddingsModel and LoadEmbeddingsModel:\n",
+    "  ECommand = \"--embeddingsmodel emodel.bin\"\n",
+    "else:\n",
+    "  ECommand = \"\"\n",
     "if FlashAttention:\n",
     "  FACommand = \"--flashattention\"\n",
     "else:\n",
@@ -152,6 +160,8 @@
     "if TTSCommand:\n",
     "  !aria2c -x 10 -o ttsmodel.bin --summary-interval=5 --download-result=default --allow-overwrite=true --file-allocation=none $TTSModel\n",
     "  !aria2c -x 10 -o ttswavtok.bin --summary-interval=5 --download-result=default --allow-overwrite=true --file-allocation=none $WavTokModel\n",
+    "if ECommand:\n",
+    "  !aria2c -x 10 -o emodel.bin --summary-interval=5 --download-result=default --allow-overwrite=true --file-allocation=none $EmbeddingsModel\n",
     "\n",
     "if MakeLocalTunnelFallback:\n",
     "  import urllib\n",
@@ -165,7 +175,7 @@
     "  print(f\"Please open the above link, and input the password '{ltpw}'\\nYour KoboldCpp will start shortly...\")\n",
     "  print(\"=================\")\n",
     "  !sleep 10\n",
-    "!./koboldcpp_linux model.gguf --usecublas 0 mmq --chatcompletionsadapter AutoGuess --multiuser --gpulayers $Layers --contextsize $ContextSize --websearch --quiet --remotetunnel $FACommand $MPCommand $VCommand $SCommand $WCommand $TTSCommand $SavGdriveCommand\n"
+    "!./koboldcpp_linux model.gguf --usecublas 0 mmq --chatcompletionsadapter AutoGuess --multiuser --gpulayers $Layers --contextsize $ContextSize --websearch --quiet --remotetunnel $FACommand $MPCommand $VCommand $SCommand $WCommand $TTSCommand $ECommand $SavGdriveCommand\n"
    ]
   }
  ],