Setting near and far with scene bounds collider. Also LR schedules. (#74)

* some initial cuda code as I learn. using git to save progress.

* first cuda kernel for ray sampling that mostly works

* cleaning up code. working on lr. structs for scene bounds aabb

* added scene bounds collider code

* updated setup

* setup

* setup

* black

* including some files

* exporting env variable

* new file

* git workflows

* github actions

* github actions

* github actions

* removing env var

scripts/benchmarking/run_benchmark.py

@@ -56,7 +56,7 @@ def main():
         trainer = Trainer(config, local_rank=0, world_size=1)
         trainer.setup(test_mode=True)
         avg_psnr = 0
-        for step, image_idx in enumerate(config.data.validation_image_indices):
+        for step, image_idx in enumerate(config.data.val_image_indices):
             with torch.no_grad():
                 psnr = trainer.test_image(image_idx=image_idx, step=step)
             avg_psnr = (step * avg_psnr + psnr) / (step + 1)

scripts/debugging/run_actions.py

@@ -15,8 +15,8 @@ if __name__ == "__main__":
             if step["name"] == "Run Black":
                 curr_command = f"black mattport/ {step['with']['options']};"
             else:
-                compressed = step["run"].replace("\n", "").replace("\\", "")
-                curr_command = f"{compressed};"
+                compressed = step["run"].replace("\n", ";").replace("\\", "")
+                curr_command = f"{compressed}"
 
             print("*" * 100)
             print(f"Running {curr_command}")

scripts/run_train.py

@@ -18,6 +18,7 @@ from omegaconf import DictConfig
 
 from mattport.nerf.trainer import Trainer
 from mattport.utils import comms, profiler
+from time import time
 
 logging.basicConfig(format="[%(filename)s:%(lineno)d] %(message)s", level=logging.DEBUG)
 
@@ -99,7 +100,7 @@ def _distributed_worker(
 
     assert num_gpus_per_machine <= torch.cuda.device_count()
     torch.cuda.set_device(local_rank)
-    _set_random_seed(1234 + local_rank)  # NOTE(ethan): will this work for multiple machines? should we use global_rank?
+    _set_random_seed(1234 + global_rank)
     comms.synchronize(world_size)
 
     output = main_func(local_rank, world_size, config)
@@ -148,9 +149,11 @@ def launch(
     world_size = num_machines * num_gpus_per_machine
     if world_size == 0:
         # Using only CPU and one process.
+        _set_random_seed(0)
         main_func(local_rank=0, world_size=0, config=config)
     elif world_size == 1:
         # Using one gpu and one process.
+        _set_random_seed(0)
         try:
             main_func(local_rank=0, world_size=1, config=config)
         except KeyboardInterrupt: