finish TrustRegionpolicyOptimizationLoss #113

S1/uucoco_#113/TrustRegionPolicyOptimizationLoss_cuda.py

@@ -0,0 +1,59 @@
+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 trpo_kernel(const float* log_probs, const float* old_log_probs, const float* advantages, const float* old_probs, const float* new_probs, float* surrogate_out, float* kl_out, int batch_size, int action_dim) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (idx < batch_size) {
+        float ratio = expf(log_probs[idx] - old_log_probs[idx]);
+        surrogate_out[idx] = -ratio * advantages[idx];
+
+        float kl_sum = 0.0f;
+        for (int i = 0; i < action_dim; i++) {
+            float old_p = old_probs[idx * action_dim + i];
+            float new_p = new_probs[idx * action_dim + i];
+            if (old_p > 1e-8f && new_p > 1e-8f) {
+                kl_sum += old_p * (logf(old_p) - logf(new_p));
+            }
+        }
+        kl_out[idx] = kl_sum;
+    }
+}
+
+torch::Tensor trpo_cuda(torch::Tensor log_probs, torch::Tensor old_log_probs, torch::Tensor advantages, torch::Tensor old_probs, torch::Tensor new_probs, float max_kl) {
+    auto batch_size = log_probs.size(0);
+    auto action_dim = old_probs.size(1);
+    auto surrogate_out = torch::empty({batch_size}, log_probs.options());
+    auto kl_out = torch::empty({batch_size}, log_probs.options());
+    const int block_size = 256;
+    int num_blocks = (batch_size + block_size - 1) / block_size;
+    trpo_kernel<<<num_blocks, block_size>>>(log_probs.data_ptr<float>(), old_log_probs.data_ptr<float>(), advantages.data_ptr<float>(), old_probs.data_ptr<float>(), new_probs.data_ptr<float>(), surrogate_out.data_ptr<float>(), kl_out.data_ptr<float>(), batch_size, action_dim);
+    return surrogate_out.mean() + max_kl * kl_out.mean();
+}
+"""
+
+cpp_source = """
+torch::Tensor trpo_cuda(torch::Tensor log_probs, torch::Tensor old_log_probs, torch::Tensor advantages, torch::Tensor old_probs, torch::Tensor new_probs, float max_kl);
+"""
+
+trpo_module = load_inline(
+    name="trpo_loss",
+    cpp_sources=cpp_source,
+    cuda_sources=cuda_source,
+    functions=["trpo_cuda"],
+    verbose=True
+)
+
+
+class ModelNew(torch.nn.Module):
+    def __init__(self, max_kl):
+        super(ModelNew, self).__init__()
+        self.max_kl = max_kl
+        self.trpo_module = trpo_module
+
+    def forward(self, log_probs, old_log_probs, advantages, old_probs, new_probs):
+        return self.trpo_module.trpo_cuda(log_probs, old_log_probs, advantages, old_probs, new_probs, self.max_kl)

S1/uucoco_#113/TrustRegionPolicyOptimizationLoss_torch.py

@@ -0,0 +1,34 @@
+import torch
+import torch.nn as nn
+
+
+class Model(nn.Module):
+    def __init__(self, max_kl):
+        super(Model, self).__init__()
+        self.max_kl = max_kl
+
+    def forward(self, log_probs: torch.Tensor, old_log_probs: torch.Tensor, advantages: torch.Tensor,
+                old_probs: torch.Tensor, new_probs: torch.Tensor) -> torch.Tensor:
+        ratio = torch.exp(log_probs - old_log_probs)
+        surrogate_loss = -(ratio * advantages).mean()
+        kl_div = (old_probs * (torch.log(old_probs) - torch.log(new_probs))).sum(dim=-1).mean()
+        loss = surrogate_loss + self.max_kl * kl_div
+        return loss
+
+
+batch_size = 32
+action_dim = 4
+
+
+def get_inputs():
+    log_probs = torch.randn(batch_size)
+    old_log_probs = torch.randn(batch_size)
+    advantages = torch.randn(batch_size)
+    old_probs = torch.softmax(torch.randn(batch_size, action_dim), dim=-1)
+    new_probs = torch.softmax(torch.randn(batch_size, action_dim), dim=-1)
+    return [log_probs, old_log_probs, advantages, old_probs, new_probs]
+
+
+def get_init_inputs():
+    max_kl = torch.tensor(0.01)
+    return [max_kl]

S1/uucoco_#113/prompt.txt

@@ -0,0 +1,59 @@
+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.
+PyTorch C++/CUDA Extension: Inline compilation using torch.utils.cpp_extension.load_inline.
+
+Custom TRPO Loss Kernel: Computes two components per batch item:
+
+Surrogate objective: -ratio * advantages, where ratio = exp(log_probs - old_log_probs).
+
+KL divergence: Σ old_p * (log(old_p) - log(new_p)) across action dimensions.
+
+Serial KL Summation: Each thread loops over all action dimensions to compute the KL term.
+
+Fixed Block Configuration: 256 threads per block, grid size based on batch size.
+
+Regularized Loss Combination: Returns surrogate_loss + max_kl * kl_loss after taking means.
+
+Numerical Safety: Checks probability values > 1e-8 before log operations.
+
+Python Wrapper with Hyperparameter: Accepts max_kl as a constructor argument, passed to the CUDA function.
+
+
+
+
+
+
+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, max_kl):
+        super(Model, self).__init__()
+        self.max_kl = max_kl
+
+    def forward(self, log_probs: torch.Tensor, old_log_probs: torch.Tensor, advantages: torch.Tensor,
+                old_probs: torch.Tensor, new_probs: torch.Tensor) -> torch.Tensor:
+        ratio = torch.exp(log_probs - old_log_probs)
+        surrogate_loss = -(ratio * advantages).mean()
+        kl_div = (old_probs * (torch.log(old_probs) - torch.log(new_probs))).sum(dim=-1).mean()
+        loss = surrogate_loss + self.max_kl * kl_div
+        return loss
+
+
+batch_size = 32
+action_dim = 4
+
+
+def get_inputs():
+    log_probs = torch.randn(batch_size)
+    old_log_probs = torch.randn(batch_size)
+    advantages = torch.randn(batch_size)
+    old_probs = torch.softmax(torch.randn(batch_size, action_dim), dim=-1)
+    new_probs = torch.softmax(torch.randn(batch_size, action_dim), dim=-1)
+    return [log_probs, old_log_probs, advantages, old_probs, new_probs]
+
+
+def get_init_inputs():
+    max_kl = torch.tensor(0.01)
+    return [max_kl]

S1/uucoco_#113/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 TrustRegionPolicyOptimizationLoss_torch import Model, get_inputs, get_init_inputs
+from TrustRegionPolicyOptimizationLoss_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()