Upload 8 files

Dockerfile-llama-cpp-wheel

@@ -1,6 +1,6 @@
 FROM nvidia/cuda:11.3.1-cudnn8-devel-ubuntu18.04
 
-ARG LLAMA_CPP_VERSION="0.1.52"
+ARG LLAMA_CPP_VERSION="0.1.53"
 ARG CMAKE_VERSION=3.26
 ARG CMAKE_VERSION_PATCH=3.26.3
 ARG CMAKE_OS=linux

README.md

@@ -1,8 +1,8 @@
 ---
-title: Vicuna V1.3 GGML
-emoji: 🏃
-colorFrom: blue
-colorTo: gray
+title: Llama 2 13B Novel French GGML
+emoji: 📖📚
+colorFrom: #0055A4
+colorTo: #EF3340
 sdk: gradio
 sdk_version: 3.29.0
 app_file: tabbed.py

config.yml

@@ -1,11 +1,9 @@
 ---
 hub:
-  repo_id: TheBloke/vicuna-13b-v1.3.0-GGML
-  filename: vicuna-13b-v1.3.0.ggmlv3.q2_K.bin
-  # repo_id: TheBloke/Wizard-Vicuna-13B-Uncensored-GGML
-  # filename: Wizard-Vicuna-13B-Uncensored.ggmlv3.q4_1.bin
+  repo_id: Nekochu/Llama-2-13B-fp16-french
+  filename: llama-2-13b-chat-fp16FR.ggmlv3.q4_K_M.bin
 llama_cpp:
-  n_ctx: 2048
+  n_ctx: 4096
   # n_gpu_layers: 40  # llama 13b has 40 layers
 chat:
   stop:

tabbed.py

@@ -114,19 +114,20 @@ instruct_interface = gr.Interface(
     outputs=gr.outputs.Textbox(label="Output text"),
 )
 
+
 with gr.Blocks() as demo:
     with gr.Row():
         with gr.Column():
             gr.Markdown(f"""
-                    ### brought to you by OpenAccess AI Collective
-                    - Unquantized model available at https://huggingface.co/openaccess-ai-collective/manticore-13b-chat-pyg
-                    - This is the [{config["hub"]["repo_id"]}](https://huggingface.co/{config["hub"]["repo_id"]}) model file [{config["hub"]["filename"]}](https://huggingface.co/{config["hub"]["repo_id"]}/blob/main/{config["hub"]["filename"]})
-                    - This Space uses GGML with GPU support, so it can quickly run larger models on smaller GPUs & VRAM.
-                    - This is running on a smaller, shared GPU, so it may take a few seconds to respond. 
-                    - [Duplicate the Space](https://huggingface.co/spaces/openaccess-ai-collective/ggml-ui?duplicate=true) to skip the queue and run in a private space or to use your own GGML models.
-                    - When using your own models, simply update the [config.yml](https://huggingface.co/spaces/openaccess-ai-collective/ggml-ui/blob/main/config.yml)
-                    - Contribute at [https://github.com/OpenAccess-AI-Collective/ggml-webui](https://github.com/OpenAccess-AI-Collective/ggml-webui)
-                    - Many thanks to [TheBloke](https://huggingface.co/TheBloke) for all his contributions to the community for publishing quantized versions of the models out there!  
+                    ### This is the [{config["hub"]["repo_id"]}](https://huggingface.co/{config["hub"]["repo_id"]}) quantized model file [{config["hub"]["filename"]}](https://huggingface.co/{config["hub"]["repo_id"]}/blob/main/{config["hub"]["filename"]})
+    
+                    <details>
+                        <summary><a href="https://huggingface.co/spaces/openaccess-ai-collective/ggml-ui?duplicate=true">Duplicate the Space</a> to skip the queue and run in a private space or to use your own GGML models, simply update the <a href="https://huggingface.co/spaces/openaccess-ai-collective/ggml-ui/blob/main/config.yml">config.yml</a></summary>
+                        <ul>
+                            <li>This Space uses GGML with GPU support, so it can quickly run larger models on smaller GPUs & VRAM. [<a href="https://github.com/OpenAccess-AI-Collective/ggml-webui">Contribute</a>]</li>
+                            <li>This is running on a smaller, shared GPU, so it may take a few seconds to respond.</li>
+                        </ul>
+                    </details>
                     """)
     with gr.Tab("Chatbot"):
         gr.Markdown("# GGML Spaces Chatbot Demo")
@@ -177,4 +178,4 @@ with gr.Blocks() as demo:
         gr.Markdown("# GGML Spaces Instruct Demo")
         instruct_interface.render()
 
-demo.queue(**config["queue"]).launch(debug=True, server_name="0.0.0.0", server_port=7860)
+demo.queue(**config["queue"]).launch(debug=True, server_name="0.0.0.0", server_port=7860)