#!/usr/bin/env bash
# Hydrogen-1.0 lokal ausprobieren – Linux
# Automatisch erzeugt von Tokenstrom (tokenstrom.com).
# Aufruf: bash try_linux.sh
set -e
cd "$(dirname "$0")"
echo "== Hydrogen-1.0 =="
if command -v nvidia-smi >/dev/null 2>&1; then nvidia-smi --query-gpu=name,memory.total --format=csv,noheader;
else echo "No NVIDIA GPU found – running on CPU (slow)"; fi
echo "WARNUNG: Groß (753,3 B parameters, MoE): running entirely on the GPU takes a lot of VRAM (video memory). With MoE offloading, a regular graphics card plus plenty of RAM is enough (see table). Colab Free isn't enough."

echo "== Python / Hugging Face =="
python3 -m venv .venv-hydrogen-1-0
source .venv-hydrogen-1-0/bin/activate
pip install -q -U pip
if command -v nvidia-smi >/dev/null 2>&1; then
  pip install -q -U torch --index-url https://download.pytorch.org/whl/cu124
  pip install -q -U transformers accelerate bitsandbytes
else
  pip install -q -U torch --index-url https://download.pytorch.org/whl/cpu
  pip install -q -U transformers accelerate
fi
echo 'Gated model: if the download fails, run  hf auth login  first.'
python3 demo.py