forked from ccf-ai-infra/GPUCodeForces
53 lines
2.0 KiB
Python
53 lines
2.0 KiB
Python
import torch
|
||
import time
|
||
from torchcode import Model, get_inputs, get_init_inputs
|
||
from cudacode import ModelNew
|
||
|
||
def run_benchmark():
|
||
if not torch.cuda.is_available():
|
||
print("CUDA 不可用，请确保您有可用的 NVIDIA GPU 并已正确安装 PyTorch CUDA 版本。")
|
||
return
|
||
device = torch.device("cuda")
|
||
|
||
init_inputs = [x.cuda(device=device) if isinstance(x, torch.Tensor) else x for x in get_init_inputs()]
|
||
inputs = [x.cuda(device=device) if isinstance(x, torch.Tensor) else x for x in get_inputs()]
|
||
|
||
torch_model = Model(*init_inputs).cuda()
|
||
cuda_model = ModelNew(*init_inputs).cuda()
|
||
torch_model.eval(); cuda_model.eval()
|
||
|
||
print("-------------------- 精度对齐验证 --------------------")
|
||
with torch.no_grad():
|
||
out_torch = torch_model(*inputs)
|
||
out_cuda = cuda_model(*inputs)
|
||
flag = torch.allclose(out_torch, out_cuda, rtol=1e-03)
|
||
if flag:
|
||
print("✅ 精度对齐：两个模型的输出结果非常接近。")
|
||
else:
|
||
print("❌ 精度不一致！")
|
||
print(f"最大绝对误差: {(out_torch - out_cuda).abs().max().item()}" )
|
||
|
||
print("\n-------------------- 性能加速比测试 --------------------")
|
||
iters = 50
|
||
torch.cuda.synchronize(); t0 = time.time()
|
||
for _ in range(iters):
|
||
_ = torch_model(*inputs)
|
||
torch.cuda.synchronize(); t_torch = (time.time() - t0) / iters
|
||
|
||
torch.cuda.synchronize(); t0 = time.time()
|
||
for _ in range(iters):
|
||
_ = cuda_model(*inputs)
|
||
torch.cuda.synchronize(); t_cuda = (time.time() - t0) / iters
|
||
|
||
print(f"PyTorch Spatial-Diff-Gate 平均执行时间: {t_torch:.6f} 秒")
|
||
print(f"自定义 CUDA 融合内核 平均执行时间: {t_cuda:.6f} 秒")
|
||
sp = t_torch / t_cuda if t_cuda > 0 else 0
|
||
if t_cuda > 0:
|
||
print(f"加速比 (Speedup): {sp:.2f}x")
|
||
else:
|
||
print("CUDA 内核执行时间为0，无法计算加速比。")
|
||
return flag, sp
|
||
|
||
if __name__ == "__main__":
|
||
run_benchmark()
|