width/height mismatch fix, cleanup

model.cpp

@@ -127,7 +127,7 @@ torch::Tensor Model::forward(Camera& cam, int step){
         conics = p[3];
         numTilesHit = p[4];
         #else
-            throw std::runtime_error("GPU support not built");
+            throw std::runtime_error("GPU support not built, use --cpu");
         #endif
     }
 
@@ -143,7 +143,6 @@ torch::Tensor Model::forward(Camera& cam, int step){
     int degreesToUse = (std::min<int>)(step / shDegreeInterval, shDegree);
     torch::Tensor rgbs;
 
-    std::cerr << "HERE";
     if (device == torch::kCPU){
         rgbs = SphericalHarmonicsCPU::apply(degreesToUse, viewDirs, colors);
     }else{
@@ -152,7 +151,6 @@ torch::Tensor Model::forward(Camera& cam, int step){
         #endif
     }
 
-    std::cerr << "THERE" << camDepths;
     rgbs = torch::clamp_min(rgbs + 0.5f, 0.0f);
 
 

opensplat.cpp

@@ -77,10 +77,12 @@ int main(int argc, char *argv[]){
     const float splitScreenSize = result["split-screen-size"].as<float>();
 
     torch::Device device = torch::kCPU;
+    int displayStep = 1;
 
     if (torch::cuda::is_available() && result.count("cpu") == 0) {
         std::cout << "Using CUDA" << std::endl;
         device = torch::kCUDA;
+        displayStep = 10;
     }else{
         std::cout << "Using CPU" << std::endl;
     }
@@ -120,7 +122,7 @@ int main(int argc, char *argv[]){
             torch::Tensor mainLoss = model.mainLoss(rgb, gt, ssimWeight);
             mainLoss.backward();
 
-            if (step % 10 == 0) std::cout << "Step " << step << ": " << mainLoss.item<float>() << std::endl;
+            if (step % displayStep == 0) std::cout << "Step " << step << ": " << mainLoss.item<float>() << std::endl;
 
             model.optimizersStep();
             model.schedulersStep(step);

simple_trainer.cpp

@@ -147,7 +147,7 @@ int main(int argc, char **argv){
                                 height / 2,
                                 height,
                                 width);
-
+            
             outImg = RasterizeGaussiansCPU::apply(
                 p[0], // xys
                 p[1], // radii,
@@ -182,7 +182,7 @@ int main(int argc, char **argv){
                     width,
                     background);
             #else
-                throw std::runtime_error("GPU support not built");
+                throw std::runtime_error("GPU support not built, use --cpu");
             #endif
         }
 

vendor/gsplat-cpu/gsplat_cpu.cpp

@@ -240,8 +240,8 @@ std::tuple<
     }
 
     // Background
-    for (int i = 0; i < width; i++){
-        for (int j = 0; j < height; j++){
+    for (int i = 0; i < height; i++){
+        for (int j = 0; j < width; j++){
             size_t pixIdx = (i * width + j);
             float T = pFinalTs[pixIdx];