LightGlueProject/LightGlue_Deployment/test_int8_compilation.py

#!/usr/bin/env python3
"""
测试 INT8 编译是否正常工作
"""

import torch
import sys

print("Testing INT8 compilation...")
print("="*60)

# 检查CUDA
if not torch.cuda.is_available():
    print("[ERROR] CUDA not available")
    sys.exit(1)

print(f"CUDA available: {torch.cuda.is_available()}")
print(f"GPU: {torch.cuda.get_device_name(0)}")

# 检查TensorRT
try:
    import torch_tensorrt
    print(f"torch-tensorrt version: {torch_tensorrt.__version__}")
except ImportError:
    print("[ERROR] torch-tensorrt not installed")
    sys.exit(1)

# 创建简单模型测试INT8编译
print("\nCreating test model...")
class SimpleModel(torch.nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = torch.nn.Conv2d(1, 32, 3, padding=1)
        self.conv2 = torch.nn.Conv2d(32, 64, 3, padding=1)
        self.pool = torch.nn.MaxPool2d(2)
    
    def forward(self, x):
        x = torch.relu(self.conv1(x))
        x = self.pool(x)
        x = torch.relu(self.conv2(x))
        return x

model = SimpleModel().eval().cuda()
example_input = torch.randn(1, 1, 480, 640).cuda()

print("Model created successfully")
print(f"Input shape: {example_input.shape}")

# 测试INT8编译
print("\n" + "="*60)
print("Attempting INT8 compilation...")
print("WARNING: This may take 5-10 minutes")
print("="*60)
sys.stdout.flush()

try:
    print("Starting compilation...")
    sys.stdout.flush()
    
    trt_model = torch_tensorrt.compile(
        model,
        inputs=[example_input],
        enabled_precisions={torch.float, torch.int8},
        workspace_size=1 << 30,  # 1GB
        min_block_size=7,
        ir="torchscript",
    )
    
    print("\n[OK] INT8 compilation successful!")
    print("="*60)
    
    # 测试推理
    print("Testing inference...")
    with torch.no_grad():
        output = trt_model(example_input)
    print(f"Output shape: {output.shape}")
    print("[OK] Inference successful!")
    
except Exception as e:
    print(f"\n[ERROR] INT8 compilation failed: {e}")
    print("="*60)
    print("Full traceback:")
    import traceback
    traceback.print_exc()
    print("\nRecommendation: Use FP16 instead of INT8")
    sys.exit(1)

print("\n" + "="*60)
print("INT8 compilation test PASSED!")
print("You can use --tensorrt_precision int8 in your demo script")
print("="*60)