File size: 4,183 Bytes
3dbff85 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 | #!/usr/bin/env sh
# Entrypoint for the personal HF Space (msradam/riprap-nyc) on L4.
#
# Boots three things in order:
# 1. Ollama serve → granite4.1:8b on localhost:11434
# 2. riprap-models → Prithvi/TerraMind/TTM/GLiNER/Embedding on :7861
# 3. web.main → FastAPI + SSE on :7860 (HF Spaces public port)
#
# The 8B is baked into the image (see Dockerfile.l4); the EO toolchain
# (terratorch + deps) installs at runtime to keep the build sandbox
# under its disk threshold. ~2 minutes on first cold start; cached
# thereafter for the lifetime of the image.
set -e
# --- 1. EO toolchain (runtime-installed; same pattern as the canonical
# entrypoint.sh) -------------------------------------------------
EO_DIR="$HOME/.eo-pkgs"
EO_MARKER="$EO_DIR/.installed"
if [ ! -f "$EO_MARKER" ]; then
echo "[entrypoint.l4] installing EO toolchain into $EO_DIR ..."
mkdir -p "$EO_DIR"
# torchvision is now baked into the base image (Dockerfile.l4) so
# don't re-install it here — the EO_DIR shadowing copy was the
# source of the `torchvision::nms does not exist` runtime error.
if pip install --no-cache-dir --no-deps --target="$EO_DIR" \
terratorch==1.1rc6 einops diffusers timm; then
if PYTHONPATH="$EO_DIR:$PYTHONPATH" python -c "
import terratorch
import terratorch.models.backbones.terramind.model.terramind_register
from terratorch.registry import FULL_MODEL_REGISTRY
n = len([k for k in FULL_MODEL_REGISTRY if 'terramind' in k.lower()])
assert n > 0
print(f'[entrypoint.l4] terratorch ok ({n} terramind entries)')
"; then
touch "$EO_MARKER"
echo "[entrypoint.l4] EO toolchain READY"
else
echo "[entrypoint.l4] EO verify FAILED — Prithvi/TerraMind probes will skip"
fi
else
echo "[entrypoint.l4] pip install FAILED — Prithvi/TerraMind probes will skip"
fi
else
echo "[entrypoint.l4] EO toolchain already installed (cached)"
fi
export PYTHONPATH="$EO_DIR:$PYTHONPATH"
# --- 2. Ollama serve --------------------------------------------------
LOG_OLLAMA="$HOME/ollama.log"
ollama serve 2>&1 | tee "$LOG_OLLAMA" &
OLLAMA_PID=$!
for i in $(seq 1 60); do
if curl -sf http://127.0.0.1:11434/ > /dev/null 2>&1; then
echo "[entrypoint.l4] ollama up (pid $OLLAMA_PID) after ${i}s"
break
fi
if ! kill -0 "$OLLAMA_PID" 2>/dev/null; then
echo "[entrypoint.l4] FATAL: ollama serve died"
tail -40 "$LOG_OLLAMA" || true
exit 1
fi
sleep 1
done
# Granite 4.1:8b is baked. Pre-warm into VRAM so the first reconcile
# doesn't pay the ~30s model-load tax.
echo "[entrypoint.l4] pre-warming granite4.1:8b into VRAM ..."
curl -s -X POST http://127.0.0.1:11434/api/generate \
-d '{"model":"granite4.1:8b","prompt":"hi","stream":false,"keep_alive":"24h","options":{"num_predict":1}}' \
-o /dev/null --max-time 120 \
&& echo "[entrypoint.l4] granite warm" \
|| echo "[entrypoint.l4] WARNING: granite warmup failed (will load lazily)"
# --- 3. riprap-models on :7861 ---------------------------------------
# Same FastAPI app the AMD droplet runs, just rehosted in-process here
# so app/inference.py's RIPRAP_ML_BASE_URL points at localhost.
LOG_MODELS="$HOME/riprap-models.log"
uvicorn riprap_models:app --host 127.0.0.1 --port 7861 --log-level info \
> "$LOG_MODELS" 2>&1 &
MODELS_PID=$!
for i in $(seq 1 60); do
if curl -sf http://127.0.0.1:7861/healthz > /dev/null 2>&1; then
echo "[entrypoint.l4] riprap-models up (pid $MODELS_PID) after ${i}s"
break
fi
if ! kill -0 "$MODELS_PID" 2>/dev/null; then
echo "[entrypoint.l4] FATAL: riprap-models died"
tail -40 "$LOG_MODELS" || true
exit 1
fi
sleep 1
done
# --- GPU sanity --------------------------------------------------------
if command -v nvidia-smi > /dev/null 2>&1; then
echo "[entrypoint.l4] nvidia-smi:"
nvidia-smi -L || true
else
echo "[entrypoint.l4] WARNING: nvidia-smi missing — running on CPU"
fi
# --- 4. Web app (foreground) -----------------------------------------
exec uvicorn web.main:app --host 0.0.0.0 --port 7860 --log-level info
|