126 lines
3.9 KiB
Bash
Executable File
126 lines
3.9 KiB
Bash
Executable File
#!/bin/bash
|
||
# Phase 4B RMT-PPAD Epoch 1 评估测试脚本 - 验证修复后的配置
|
||
|
||
set -e
|
||
|
||
# ✅ 关键: 设置环境变量 (参考成功脚本)
|
||
export PATH=/opt/conda/bin:$PATH
|
||
export LD_LIBRARY_PATH=/opt/conda/lib/python3.8/site-packages/torch/lib:/opt/conda/lib:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
|
||
export PYTHONPATH=/workspace/bevfusion:$PYTHONPATH
|
||
#export PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:64
|
||
|
||
# 设置GPU
|
||
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
|
||
|
||
cd /workspace/bevfusion
|
||
|
||
echo "========================================================================"
|
||
echo "Phase 4B RMT-PPAD Epoch 1 评估测试"
|
||
echo "========================================================================"
|
||
echo "目的: 验证修复后的test_pipeline是否正常工作"
|
||
echo "Checkpoint: runs/run-4c8ec7e5-fabdc997/epoch_1.pth (最新)"
|
||
echo "配置: multitask_BEV2X_phase4b_rmtppad_segmentation.yaml"
|
||
echo "========================================================================"
|
||
echo ""
|
||
|
||
# 验证环境
|
||
echo ""
|
||
echo "=== 环境验证 ==="
|
||
/opt/conda/bin/python -c "import torch; print('✅ PyTorch:', torch.__version__)" || {
|
||
echo "❌ PyTorch导入失败"
|
||
exit 1
|
||
}
|
||
/opt/conda/bin/python -c "import mmcv; print('✅ mmcv:', mmcv.__version__)" || {
|
||
echo "❌ mmcv导入失败"
|
||
exit 1
|
||
}
|
||
which torchpack || {
|
||
echo "❌ torchpack未找到"
|
||
exit 1
|
||
}
|
||
echo "✅ torchpack: $(which torchpack)"
|
||
|
||
# 创建评估输出目录
|
||
EVAL_DIR="eval_test/epoch1_test_$(date +%Y%m%d_%H%M%S)"
|
||
mkdir -p "$EVAL_DIR"
|
||
|
||
CONFIG="runs/run-4c8ec7e5-fabdc997/configs.yaml"
|
||
CHECKPOINT="runs/run-4c8ec7e5-fabdc997/epoch_1.pth"
|
||
|
||
echo "配置文件: $CONFIG"
|
||
echo "Checkpoint: $CHECKPOINT"
|
||
echo "输出目录: $EVAL_DIR"
|
||
echo ""
|
||
|
||
# 检查文件存在
|
||
if [ ! -f "$CONFIG" ]; then
|
||
echo "❌ 配置文件不存在: $CONFIG"
|
||
exit 1
|
||
fi
|
||
|
||
if [ ! -f "$CHECKPOINT" ]; then
|
||
echo "❌ Checkpoint不存在: $CHECKPOINT"
|
||
exit 1
|
||
fi
|
||
|
||
echo "✓ 文件检查通过"
|
||
echo ""
|
||
|
||
# 检查GPU
|
||
GPU_COUNT=$(python -c "import torch; print(torch.cuda.device_count())" 2>/dev/null || echo "0")
|
||
echo "可用GPU数量: $GPU_COUNT"
|
||
|
||
if [ "$GPU_COUNT" -eq 0 ]; then
|
||
echo "⚠️ 没有GPU，使用CPU模式"
|
||
GPU_COUNT=1
|
||
fi
|
||
|
||
echo ""
|
||
|
||
echo "开始评估测试..."
|
||
echo "测试模式: 仅处理前10个样本 (data.samples_per_gpu=1, 总共10个batch)"
|
||
echo "预计时间: 5-10分钟"
|
||
echo "日志文件: $EVAL_DIR/eval_test.log"
|
||
echo ""
|
||
|
||
# 只测试前10个样本，验证pipeline是否工作
|
||
# 使用torchpack分布式运行 (参考训练脚本)
|
||
torchpack dist-run \
|
||
-np 1 \
|
||
/opt/conda/bin/python tools/test.py \
|
||
"$CONFIG" \
|
||
"$CHECKPOINT" \
|
||
--eval bbox map \
|
||
--out "$EVAL_DIR/test_results.pkl" \
|
||
--cfg-options data.test.samples_per_gpu=1 data.workers_per_gpu=0 \
|
||
2>&1 | tee "$EVAL_DIR/eval_test.log" | head -50
|
||
|
||
echo ""
|
||
echo "========================================================================"
|
||
echo "评估测试完成！"
|
||
echo "========================================================================"
|
||
echo "结果文件: $EVAL_DIR/test_results.pkl"
|
||
echo "日志文件: $EVAL_DIR/eval_test.log"
|
||
echo ""
|
||
|
||
# 检查是否成功完成
|
||
if grep -q "Evaluation results" "$EVAL_DIR/eval_test.log"; then
|
||
echo "✅ 评估成功！test_pipeline修复有效"
|
||
echo ""
|
||
|
||
echo "========================================================================"
|
||
echo "关键指标 (前10个样本):"
|
||
echo "========================================================================"
|
||
grep -E "(NDS|mAP|mIoU|Car|Pedestrian|Divider|Divider Dice|Heatmap Loss)" "$EVAL_DIR/eval_test.log" | grep -v "UserWarning" | tail -20
|
||
|
||
else
|
||
echo "❌ 评估失败，检查日志文件"
|
||
echo "关键错误信息:"
|
||
grep -E "(ERROR|Error|Exception|KeyError|ImportError)" "$EVAL_DIR/eval_test.log" | tail -10
|
||
exit 1
|
||
fi
|
||
|
||
echo ""
|
||
echo "如果测试通过，可以安全启动完整训练！"
|
||
echo "========================================================================"
|