more logs

Author: sebffischer

paper/benchmark/benchmark.R

@@ -2,6 +2,9 @@ library(batchtools)
 library(mlr3misc)
 
 setup = function(reg_path, python_path, work_dir) {
+
+  print_setup_info(reg_path, python_path, work_dir)
+
   reg = makeExperimentRegistry(
     file.dir = reg_path,
     work.dir = work_dir,
@@ -69,6 +72,7 @@ setup = function(reg_path, python_path, work_dir) {
   })
 
   addAlgorithm("rtorch", fun = function(instance, job, opt_type, jit, ...) {
+    print
     assert_choice(opt_type, c("standard", "ignite"))
     if (opt_type == "ignite") {
       instance$optimizer = paste0("ignite_", instance$optimizer)
@@ -94,3 +98,29 @@ REPLS = 10L
 EPOCHS = 20L
 N = 2000L
 P = 1000L
+
+print_setup_info = function(reg_path, python_path, work_dir) {
+  cat("Session Info:\n")
+  print(sessionInfo())
+  cat("Library Paths:\n")
+  for (path in .libPaths()) {
+    cat("  -", path, "\n")
+  }
+  cat("Working Directory:", getwd(), "\n")
+
+  cat("Subfolders of working directory:\n")
+  for (folder in list.files(work_dir)) {
+    cat("  -", folder, "\n")
+  }
+
+  # Function arguments
+  cat("--- FUNCTION ARGUMENTS ---\n")
+  cat("  Registry Path:", reg_path, "\n")
+  cat("  Python Path:", python_path, " (", if (file.exists(python_path)) "exists" else "does not exist", ")\n")
+  cat("  Work Directory:", work_dir, "\n\n")
+  cat("Cuda is available:", torch::cuda_is_available(), "\n")
+  cat("Torch install path:", torch::install_path(), "\n")
+  reticulate::use_python(python_path, required = TRUE)
+  reticulate::source_python(here::here("benchmark", "time_pytorch.py"))
+  print(reticulate::py_config())
+}

paper/benchmark/time_pytorch.py

@@ -6,6 +6,17 @@
 
 # 3. Define the timing function
 def time_pytorch(epochs, batch_size, n_layers, latent, n, p, device, seed, optimizer, jit):
+    print(f"epochs: {epochs}")
+    print(f"batch_size: {batch_size}")
+    print(f"n_layers: {n_layers}")
+    print(f"latent: {latent}")
+    print(f"n: {n}")
+    print(f"p: {p}")
+    print(f"device: {device}")
+    print(f"seed: {seed}")
+    print(f"optimizer: {optimizer}")
+    print(f"jit: {jit}")
+
     torch.manual_seed(seed)
     torch.set_num_threads(1)
     # convert latentn, n_layers to int

paper/benchmark/time_rtorch.R

@@ -4,6 +4,18 @@ time_rtorch = function(epochs, batch_size, n_layers, latent, n, p, device, jit,
   torch_set_num_threads(1)
   torch_manual_seed(seed)
 
+  cat("epochs: ", epochs, "\n")
+  cat("batch_size: ", batch_size, "\n")
+  cat("n_layers: ", n_layers, "\n")
+  cat("latent: ", latent, "\n")
+  cat("n: ", n, "\n")
+  cat("p: ", p, "\n")
+  cat("device: ", device, "\n")
+  cat("jit: ", jit, "\n")
+  cat("seed: ", seed, "\n")
+  cat("optimizer: ", optimizer, "\n")
+  cat("mlr3torch: ", mlr3torch, "\n")
+
   lr = 0.0001
 
   make_network = function(p, latent, n_layers) {