Merge pull request 'finish mapLoss #160' (#884) from gsd123/GPUCodeForces:gsd160 into main

S1/gsd123_#160/mapLoss_cuda.py

@@ -0,0 +1,78 @@
+import os
+import torch
+import torch.nn as nn
+from torch.utils.cpp_extension import load_inline
+
+cuda_source = """
+#include <torch/extension.h>
+#include <cuda_runtime.h>
+
+__global__ void mapping_loss_kernel(const float* __restrict__ x, 
+                                    const float* __restrict__ y, 
+                                    float* __restrict__ out, 
+                                    int dim) {
+    int bid = blockIdx.x;
+    int tid = threadIdx.x;
+
+    const float* row_x = x + bid * dim;
+    const float* row_y = y + bid * dim;
+
+    float local_sum = 0.0f;
+
+    for (int i = tid; i < dim; i += blockDim.x) {
+        local_sum += fabsf(row_x[i] - row_y[i]);
+    }
+
+    __shared__ float s_sum[256];
+    s_sum[tid] = local_sum;
+    __syncthreads();
+
+    for (int stride = blockDim.x / 2; stride > 0; stride >>= 1) {
+        if (tid < stride) {
+            s_sum[tid] += s_sum[tid + stride];
+        }
+        __syncthreads();
+    }
+
+    if (tid == 0) {
+        // Compute mean for this sample (L1 distance)
+        out[bid] = s_sum[0] / (float)dim;
+    }
+}
+
+torch::Tensor mapping_loss_cuda(torch::Tensor x, torch::Tensor y) {
+    int batch_size = x.size(0);
+    int dim = x.size(1);
+    auto out = torch::empty({batch_size}, x.options());
+
+    mapping_loss_kernel<<<batch_size, 256>>>(
+        x.data_ptr<float>(),
+        y.data_ptr<float>(),
+        out.data_ptr<float>(),
+        dim
+    );
+
+    return out;
+}
+"""
+
+cpp_source = "torch::Tensor mapping_loss_cuda(torch::Tensor x, torch::Tensor y);"
+
+module = load_inline(
+    name="mapping_loss_ext",
+    cpp_sources=cpp_source,
+    cuda_sources=cuda_source,
+    functions=["mapping_loss_cuda"],
+    verbose=False,
+    with_cuda=True
+)
+
+
+class ModelNew(nn.Module):
+    def __init__(self):
+        super(ModelNew, self).__init__()
+        self.op = module
+
+    def forward(self, x, y):
+        batch_means = self.op.mapping_loss_cuda(x.contiguous(), y.contiguous())
+        return batch_means.mean()

S1/gsd123_#160/mapLoss_torch.py

@@ -0,0 +1,24 @@
+import torch
+import torch.nn as nn
+
+
+class Model(nn.Module):
+    def __init__(self):
+        super(Model, self).__init__()
+
+    def forward(self, x, y):
+        return torch.abs(x - y).mean()
+
+
+batch_size = 16
+input_dim = 1024
+
+
+def get_inputs():
+    x = torch.randn(batch_size, input_dim)
+    y = torch.randn(batch_size, input_dim)
+    return [x, y]
+
+
+def get_init_inputs():
+    return []

S1/gsd123_#160/prompt.txt

@@ -0,0 +1,45 @@
+You write custom CUDA kernels to replace the pytorch operators in the given GeGLU architecture to get speedups.
+
+You have complete freedom to choose the set of operators you want to replace. You may make the decision to replace some operators with custom CUDA kernels and leave others unchanged. You may replace multiple operators with custom implementations, consider operator fusion opportunities (combining multiple operators into a single kernel, for example, combining chunk+gelu+elementwise_mul), or algorithmic changes (such as optimized memory access patterns). You are only limited by your imagination.
+
+CUDA C++ kernel for element‑wise L1 mapping loss (mean absolute error)
+
+Block‑parallel per‑sample processing: each block handles one batch element
+
+Thread‑wise accumulation of absolute differences (|x[i] – y[i]|)
+
+Parallel reduction in shared memory using binary tree approach
+
+Mean computation: after reduction, divides by feature dimension dim
+
+Coalesced memory access with grid‑stride loops
+
+PyTorch inline C++/CUDA extension via load_inline
+
+
+
+Here's an example to show you the syntax of inline embedding custom CUDA operators in torch: The example given architecture is:
+import torch
+import torch.nn as nn
+
+
+class Model(nn.Module):
+    def __init__(self):
+        super(Model, self).__init__()
+
+    def forward(self, x, y):
+        return torch.abs(x - y).mean()
+
+
+batch_size = 16
+input_dim = 1024
+
+
+def get_inputs():
+    x = torch.randn(batch_size, input_dim)
+    y = torch.randn(batch_size, input_dim)
+    return [x, y]
+
+
+def get_init_inputs():
+    return []

S1/gsd123_#160/run_code.py

@@ -0,0 +1,77 @@
+###########################################################
+# 性能和精度验证程序
+###########################################################
+import torch
+import torch.nn as nn
+import time
+from mapLoss_torch import Model, get_inputs, get_init_inputs
+from mapLoss_cuda  import ModelNew
+
+
+def run_benchmark():
+    # 检查 CUDA 是否可用
+    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
+    else:
+        device = torch.device("cuda")
+
+    # 初始化模型
+    init_inputs = get_init_inputs()
+    init_inputs = [
+        x.cuda(device=device) if isinstance(x, torch.Tensor) else x for x in init_inputs
+    ]
+    inputs = get_inputs()
+    inputs = [
+        x.cuda(device=device) if isinstance(x, torch.Tensor) else x for x in 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():
+        output_torch = torch_model(*inputs)
+        output_cuda = cuda_model(*inputs)
+
+    precision_flag = torch.allclose(output_torch, output_cuda, rtol=1e-03)
+    if precision_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("\n-------------------- 性能加速比测试 --------------------")
+    num_iterations = 100
+
+    # PyTorch 模型计时
+    torch.cuda.synchronize()
+    start_time = time.time()
+    for _ in range(num_iterations):
+        _ = torch_model(*inputs)
+    torch.cuda.synchronize()
+    torch_time = (time.time() - start_time) / num_iterations
+
+    # 自定义 CUDA 内核计时
+    torch.cuda.synchronize()
+    start_time = time.time()
+    for _ in range(num_iterations):
+        _ = cuda_model(*inputs)
+    torch.cuda.synchronize()
+    cuda_time = (time.time() - start_time) / num_iterations
+
+    print(f"PyTorch torch.relu 平均执行时间: {torch_time:.6f} 秒")
+    print(f"自定义 CUDA 内核 平均执行时间: {cuda_time:.6f} 秒")
+    speedup = 0
+    if cuda_time > 0:
+        speedup = torch_time / cuda_time
+        print(f"加速比 (Speedup): {speedup:.2f}x")
+    else:
+        print("CUDA 内核执行时间为0，无法计算加速比。")
+    return precision_flag, speedup
+
+
+if __name__ == "__main__":
+    precision_flag, speedup = run_benchmark()