Upload 2 files

Dockerfile

@@ -0,0 +1,21 @@
+FROM ghcr.io/ggerganov/llama.cpp:server-cuda-7c27a19b2eb91bb0f43c7f7aec0386cec2dddc33
+
+RUN apt-get update && apt-get install -y wget pip python3 git 
+
+ENTRYPOINT ["/bin/sh", "-c"]
+
+WORKDIR llama.cpp
+
+RUN wget -q https://huggingface.co/Mihaiii/shieldgemma-2b-Q5_K_M-GGUF/resolve/main/shieldgemma-2b-q5_k_m-imat.gguf
+RUN wget -q https://huggingface.co/bartowski/gemma-2-2b-it-GGUF/resolve/main/gemma-2-2b-it-Q5_K_M.gguf
+
+COPY script.sh .
+
+RUN chmod +x ./script.sh
+RUN chmod 777 .
+
+RUN git clone https://github.com/Mihaiii/trivia.git
+RUN cd trivia && pip install -r requirements.txt
+RUN chmod 777 ./trivia
+
+CMD ["./script.sh"]

script.sh

@@ -0,0 +1,8 @@
+#!/bin/bash
+../llama-server -m ./shieldgemma-2b-q5_k_m-imat.gguf -c 8192 -t 4 --port 8000 --host 127.0.0.1 --parallel 2 -cb &
+../llama-server -m ./gemma-2-2b-it-Q5_K_M.gguf -c 8192 -t --port 7888 --host 127.0.0.1 --parallel 2 -cb &
+
+sleep 5
+
+cd trivia
+uvicorn app:app --port 7860 --host 0.0.0.0