Trying to create a runpod template for the LTX Video trainer,
I used this dockerfile for the image :
FROM pytorch/pytorch:2.2.0-cuda12.1-cudnn8-runtime
RUN apt-get update && apt-get install -y \
git curl ffmpeg build-essential \
&& rm -rf /var/lib/apt/lists/*
RUN curl -Ls https://astral.sh/uv/install.sh | bash \
&& cp /root/.local/bin/uv /usr/local/bin/uv
WORKDIR /app
RUN git clone https://github.com/Lightricks/LTX-Video-Trainer.git .
RUN uv venv && uv sync
# ← New bits:
COPY start.sh /start.sh
RUN chmod +x /start.sh
ENTRYPOINT ["/start.sh"]
the start.sh file contains this
#!/usr/bin/env bash
set -e
# 1) activate your venv
source /app/.venv/bin/activate
# 2) hand off to bash (keeps the shell open & interactive)
exec /usr/bin/bash
the docker image runs fine locally, but when running a pod with a template (shown in the screenshot) it doesn't work, no way to connect.
What did I miss? this is the first time I try make a RunPod template