#!/usr/bin/env bash
set -euo pipefail

# =========================
# Resolve paths
# =========================
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
EQUIFORMER_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
REPO_ROOT="$(cd "$SCRIPT_DIR/../../.." && pwd)"

TRAIN_SCRIPT="$EQUIFORMER_DIR/training_equiformer.py"

export PYTHONPATH="$REPO_ROOT:${PYTHONPATH:-}"

# =========================
# User settings
# =========================
NAME="equiformer"

MODEL_CONFIG="small"          # choices: orig, small
SEED="42"
DATASET="$REPO_ROOT/datasets/h5/train_plus_val.h5"


# =========================
# Detect number of visible GPUs
# =========================
if command -v nvidia-smi &> /dev/null; then
    NGPUS=$(nvidia-smi -L | wc -l)
else
    echo "ERROR: nvidia-smi not found"
    exit 1
fi

if [ "$NGPUS" -eq 0 ]; then
    echo "ERROR: No GPUs visible"
    exit 1
fi

echo "Found $NGPUS GPU(s)"
echo "CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES:-<not set>}"

echo "SCRIPT_DIR=$SCRIPT_DIR"
echo "EQUIFORMER_DIR=$EQUIFORMER_DIR"
echo "REPO_ROOT=$REPO_ROOT"

echo ""
echo "Run configuration:"
echo "  NAME=$NAME"
echo "  MODEL_CONFIG=$MODEL_CONFIG"
echo "  DATASET=$DATASET"
echo "  SEED=$SEED"
echo ""

# =========================
# Optional args
# =========================
EXTRA_ARGS=()



# =========================
# Launch with torchrun
# =========================
torchrun \
    --standalone \
    --nproc_per_node="$NGPUS" \
    "$TRAIN_SCRIPT" \
    --name "$NAME" \
    --model_config "$MODEL_CONFIG" \
    --dataset "$DATASET" \
    --seed "$SEED" \
    "${EXTRA_ARGS[@]}"