update

app.py

@@ -30,7 +30,6 @@ version_str="".join([
 ])
 # install pytorch3d with the right version
 os.system('pip install iopath')
-os.system('pip install fvcore')
 os.system(f'pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html')
 
 import pytorch3d # check whether the version matches

guided_diffusion/dist_util.py

@@ -65,7 +65,10 @@ def setup_dist(args):
 
     # dist.init_process_group(backend='nccl', init_method='env://', rank=args.local_rank, world_size=th.cuda.device_count(),  timeout=datetime.timedelta(seconds=5400))
     # st() no mark
-    dist.init_process_group(backend='nccl', init_method='env://', timeout=datetime.timedelta(seconds=54000))
+    # dist.init_process_group(backend='nccl', init_method='env://', timeout=datetime.timedelta(seconds=54000))
+    # print(f"{args.local_rank=} init complete")
+
+    dist.init_process_group(backend='gloo', init_method='env://', timeout=datetime.timedelta(seconds=54000))
     print(f"{args.local_rank=} init complete")
 
     # synchronize() # extra memory on rank 0, why?