#!/usr/bin/env bash
set -e

# 1. Rimuovi eventuali container attivi
sudo docker rm -f nim-llama3-8b-spark nim-qwen3-32b-spark 2>/dev/null || true

# 2. Estrai la chiave API
API_KEY=$(grep '^nvapi-' /home/azurian/.nvidia | head -n 1 | tr -d '\r\n[:space:]')

# 3. Assicura i permessi sulla cache
sudo chmod -R 777 /home/azurian/.cache/nim 2>/dev/null || true

# 4. Avvia il container NIM Llama 3.1 8B DGX Spark
sudo docker run -d --restart unless-stopped \
  --name nim-llama3-8b-spark \
  --gpus all \
  --ipc=host \
  -p 8001:8000 \
  -v /home/azurian/.cache/nim:/opt/nim/.cache \
  -e NGC_API_KEY="$API_KEY" \
  -e NIM_CACHE_PATH=/opt/nim/.cache \
  -e NIM_MAX_MODEL_LEN=8192 \
  -e CUDA_MODULE_LOADING=LAZY \
  -e PYTORCH_CUDA_ALLOC_CONF="expandable_segments:True" \
  nvcr.io/nim/meta/llama-3.1-8b-instruct-dgx-spark:latest

# 5. Monitora i log
sudo docker logs -f nim-llama3-8b-spark