Convert the space into a Docker space

Dockerfile-llm-app

@@ -1,18 +1,21 @@
 FROM pytorch/pytorch:2.4.0-cuda12.4-cudnn9-devel
 
-WORKDIR /app
-
 RUN apt-get update
 RUN apt-get install -y build-essential
 RUN apt-get install -y git
 
-COPY requirements.txt requirements.txt
+RUN useradd -m -u 1000 user
+USER user
+ENV HOME=/home/user \
+	PATH=/home/user/.local/bin:$PATH
+WORKDIR ${HOME}/app
+
+COPY --chown=user requirements.txt requirements.txt
 RUN pip install -r requirements.txt
-RUN CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install "llama-cpp-python<=0.2.79.0"
-COPY *.py ./
-ADD llmlib ./llmlib
+COPY --chown=user *.py ./
+ADD --chown=user llmlib ./llmlib
 RUN pip install -e llmlib
-ADD .streamlit .streamlit
+ADD --chown=user .streamlit .streamlit
 
 #CMD [ "python", "--version"]
 # CMD ["nvidia-smi"]

README.md

@@ -1,11 +1,10 @@
 ---
 title: Multimodal Vibe Check
-emoji: 👀
+emoji: 😎
 colorFrom: red
 colorTo: purple
-sdk: streamlit
-sdk_version: 1.41.1
-app_file: st_app.py
+sdk: docker
+app_port: 8020
 pinned: false
 ---
 

docker-compose.yml

@@ -3,7 +3,7 @@ x-common-gpu: &common-gpu
     dockerfile: Dockerfile-llm-app
   environment:
     - OPENAI_API_KEY=${OPENAI_API_KEY}
-    - HF_HOME=/app/.cache/huggingface
+    - HF_HOME=/home/user/.cache/huggingface
     - HF_TOKEN=${HF_TOKEN}
     - LLMS_REST_API_KEY=${LLMS_REST_API_KEY}
     - BUGSNAG_API_KEY=${BUGSNAG_API_KEY}
@@ -15,7 +15,7 @@ x-common-gpu: &common-gpu
             count: all
             capabilities: [gpu]
   volumes:
-      - /home/tomasruiz/.cache/huggingface:/app/.cache/huggingface
+      - /home/tomasruiz/.cache/huggingface:/home/user/.cache/huggingface
 
 services:
 

requirements.txt

@@ -7,3 +7,4 @@ streamlit
 bitsandbytes
 accelerate
 fastapi[standard]
+flash_attn

st_app.py

@@ -1,20 +1,6 @@
 from PIL import Image
 import streamlit as st
-
-try:
-    from llmlib.runtime import filled_model_registry
-except ModuleNotFoundError:
-    import os
-    import subprocess
-
-    os.system("pip install -e ./llmlib")
-    subprocess.run(
-        "pip install flash-attn --no-build-isolation",
-        env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
-        shell=True,
-    )
-    from llmlib.runtime import filled_model_registry
-
+from llmlib.runtime import filled_model_registry
 from llmlib.model_registry import ModelEntry, ModelRegistry
 from llmlib.base_llm import Message
 from llmlib.bundler import Bundler