finish GammmaDivergenceLoss #88

S1/uucoco_#88/GammaDivergenceLoss_cuda.py

@@ -0,0 +1,113 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch.utils.cpp_extension import load_inline
+
+cuda_source = """
+#include <torch/extension.h>
+#include <cuda_runtime.h>
+
+__global__ void gamma_divergence_kernel(
+    const float* __restrict__ p,
+    const float* __restrict__ q,
+    float* __restrict__ output,
+    int num_classes,
+    float gamma
+) {
+    extern __shared__ float sdata[];
+
+    int tid = threadIdx.x;
+    int bid = blockIdx.x; 
+
+    int row_offset = bid * num_classes;
+
+    float* s_p = sdata;
+    float* s_pq = sdata + blockDim.x;
+    float* s_q = sdata + 2 * blockDim.x;
+
+    float local_p = 0.0f;
+    float local_pq = 0.0f;
+    float local_q = 0.0f;
+
+    for (int i = tid; i < num_classes; i += blockDim.x) {
+        float p_val = p[row_offset + i];
+        float q_val = q[row_offset + i];
+
+        float q_pow_g = powf(q_val, gamma);
+
+        local_p += powf(p_val, 1.0f + gamma);
+        local_pq += p_val * q_pow_g;
+        local_q += powf(q_val, 1.0f + gamma);
+    }
+
+    s_p[tid] = local_p;
+    s_pq[tid] = local_pq;
+    s_q[tid] = local_q;
+    __syncthreads();
+
+    for (int s = blockDim.x / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            s_p[tid] += s_p[tid + s];
+            s_pq[tid] += s_pq[tid + s];
+            s_q[tid] += s_q[tid + s];
+        }
+        __syncthreads();
+    }
+
+    if (tid == 0) {
+        float sum_p = s_p[0];
+        float sum_pq = s_pq[0];
+        float sum_q = s_q[0];
+
+        float term1 = logf(sum_p) / (gamma * (1.0f + gamma));
+        float term2 = logf(sum_pq) / gamma;
+        float term3 = logf(sum_q) / (1.0f + gamma);
+
+        output[bid] = term1 - term2 + term3;
+    }
+}
+
+torch::Tensor gamma_divergence_cuda(torch::Tensor p, torch::Tensor q, float gamma) {
+    int batch_size = p.size(0);
+    int num_classes = p.size(1);
+
+    auto output = at::empty({batch_size}, p.options());
+
+    int threads = 256;
+    int blocks = batch_size;
+    int shared_mem = 3 * threads * sizeof(float);
+
+    gamma_divergence_kernel<<<blocks, threads, shared_mem>>>(
+        p.data_ptr<float>(),
+        q.data_ptr<float>(),
+        output.data_ptr<float>(),
+        num_classes,
+        gamma
+    );
+
+    return output.mean();
+}
+"""
+
+cpp_source = """
+torch::Tensor gamma_divergence_cuda(torch::Tensor p, torch::Tensor q, float gamma);
+"""
+
+gamma_divergence_loss = load_inline(
+    name="gamma_divergence_loss",
+    cpp_sources=cpp_source,
+    cuda_sources=cuda_source,
+    functions=["gamma_divergence_cuda"],
+    verbose=False
+)
+
+
+class ModelNew(nn.Module):
+    def __init__(self, gamma=0.5):
+        super(ModelNew, self).__init__()
+        self.gamma = gamma
+
+    def forward(self, p, q):
+        p_prob = F.softmax(p, dim=1)
+        q_prob = F.softmax(q, dim=1)
+        return gamma_divergence_loss.gamma_divergence_cuda(p_prob, q_prob, self.gamma)

S1/uucoco_#88/GammaDivergenceLoss_torch.py

@@ -0,0 +1,38 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class Model(nn.Module):
+    def __init__(self, gamma=0.5):
+        super(Model, self).__init__()
+        self.gamma = gamma
+
+    def forward(self, p, q):
+        p_prob = F.softmax(p, dim=1)
+        q_prob = F.softmax(q, dim=1)
+
+        sum_p_pow = torch.sum(p_prob.pow(1.0 + self.gamma), dim=1)
+        sum_pq_pow = torch.sum(p_prob * q_prob.pow(self.gamma), dim=1)
+        sum_q_pow = torch.sum(q_prob.pow(1.0 + self.gamma), dim=1)
+
+        term1 = torch.log(sum_p_pow) / (self.gamma * (1.0 + self.gamma))
+        term2 = torch.log(sum_pq_pow) / self.gamma
+        term3 = torch.log(sum_q_pow) / (1.0 + self.gamma)
+
+        loss = term1 - term2 + term3
+        return loss.mean()
+
+
+batch_size = 32
+num_classes = 1000
+
+
+def get_inputs():
+    p = torch.randn(batch_size, num_classes, requires_grad=True)
+    q = torch.randn(batch_size, num_classes)
+    return [p, q]
+
+
+def get_init_inputs():
+    return [0.5]

S1/uucoco_#88/prompt.txt

@@ -0,0 +1,124 @@
+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.
+Technologies Used in This Code
+Core Libraries & Frameworks
+PyTorch: Deep learning framework
+
+CUDA: NVIDIA's parallel computing platform for GPU acceleration
+
+C++: For high-performance kernel implementation
+
+PyTorch Specific Components
+torch.nn.Module: Base class for neural network modules
+
+torch.nn.functional.F.softmax: Softmax activation function
+
+torch.utils.cpp_extension.load_inline: For inline compilation of CUDA/C++ extensions
+
+PyTorch Tensors: Multi-dimensional arrays with automatic differentiation
+
+CUDA/C++ Implementation Details
+CUDA Kernels: Custom GPU kernel (gamma_divergence_kernel)
+
+CUDA Math Functions: powf() for exponentiation, logf() for logarithms
+
+Parallel Reduction: Tree-based reduction with multiple accumulators
+
+Shared Memory: Using __shared__ with triple-buffer pattern (s_p, s_pq, s_q)
+
+Block-Level Parallelism: One CUDA block per batch element
+
+Thread-Level Parallelism: Parallel reduction across class dimensions
+
+Mathematical Components
+Gamma Divergence: Information-geometric divergence measure
+
+Logarithmic Terms: Three logarithmic terms with different denominators
+
+Power Computations: Multiple powf() calls with (1 + gamma) exponent
+
+Normalization: Gamma parameter scaling in denominator terms
+
+Statistical Distance: Measures difference between probability distributions
+
+Memory & Parallelism Patterns
+Triple Shared Memory Buffers: Separate buffers for p, pq, and q summations
+
+Batch-Level Parallelism: Each batch element processed by separate CUDA block
+
+Class-Level Parallelism: Threads parallelize across class dimensions
+
+Hierarchical Reduction: Two-level parallel reduction within blocks
+
+Optimization Techniques
+Shared Memory Optimization: Efficient triple-buffer layout
+
+Coalesced Memory Access: Sequential memory access patterns
+
+Fused Computation: Complete divergence calculation per batch element
+
+Logarithm Post-processing: Log operations after reduction (numerically stable)
+
+Performance Features
+Massive Parallelization: GPU acceleration for divergence computation
+
+Numerical Stability: Log operations performed after summation
+
+Memory Efficiency: Shared memory reuse across multiple reductions
+
+Batch Independence: Parallel processing of batch elements
+
+Host-Device Coordination: Final mean computation on CPU
+
+Unique Implementation Aspects
+Per-Batch Block Assignment: One CUDA block per batch element
+
+Triple Reduction Pattern: Simultaneous reduction of three different sums
+
+Logarithmic Normalization: Log operations in final divergence formula
+
+Gamma Parameter Scaling: Parameter appears in all three denominator terms
+
+
+
+
+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
+import torch.nn.functional as F
+
+
+class Model(nn.Module):
+    def __init__(self, gamma=0.5):
+        super(Model, self).__init__()
+        self.gamma = gamma
+
+    def forward(self, p, q):
+        p_prob = F.softmax(p, dim=1)
+        q_prob = F.softmax(q, dim=1)
+
+        sum_p_pow = torch.sum(p_prob.pow(1.0 + self.gamma), dim=1)
+        sum_pq_pow = torch.sum(p_prob * q_prob.pow(self.gamma), dim=1)
+        sum_q_pow = torch.sum(q_prob.pow(1.0 + self.gamma), dim=1)
+
+        term1 = torch.log(sum_p_pow) / (self.gamma * (1.0 + self.gamma))
+        term2 = torch.log(sum_pq_pow) / self.gamma
+        term3 = torch.log(sum_q_pow) / (1.0 + self.gamma)
+
+        loss = term1 - term2 + term3
+        return loss.mean()
+
+
+batch_size = 32
+num_classes = 1000
+
+
+def get_inputs():
+    p = torch.randn(batch_size, num_classes, requires_grad=True)
+    q = torch.randn(batch_size, num_classes)
+    return [p, q]
+
+
+def get_init_inputs():
+    return [0.5]

S1/uucoco_#88/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 GammaDivergenceLoss_torch import Model, get_inputs, get_init_inputs
+from GammaDivergenceLoss_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()