Jinlai Zhang

bugs in distributed training

1

we are using the Slurm Workload Manager, but when compile custom operators, bugs occur: ```shell OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root....

torch default process group is not initialized

1

```shell CloserLook3D/pytorch/utils/util.py", line 49, in reduce_tensor dist.all_reduce(rt, op=dist.ReduceOp.SUM) File "/python3.6/site-packages/torch/distributed/distributed_c10d.py", line 903, in all_reduce _check_default_pg() File "/python3.6/site-packages/torch/distributed/distributed_c10d.py", line 192, in _check_default_pg "Default process group is not initialized" AssertionError: Default process...

ModuleNotFoundError: No module named 'tools.generate_detections'

2

when run detect_actions.py

: cannot connect to X server

2

when run ```shell python detect_actions.py --video_path sample_input.mp4 ```

Bugs in Build _ext module

2

``` pointnet2/_ext-src/src/group_points.cpp:56:40: error: ‘AT_CHECK’ was not declared in this scope AT_CHECK(false, "CPU not supported"); ^ error: command 'gcc' failed with exit status 1 ```

error when install

1

``` ERROR: Command errored out with exit status 128: git clone -q git://github.com/erikwijmans/etw_pytorch_utils.git /tmp/pip-install-5c0vh27l/etw-pytorch-utils_ee2fcee72ccb48a3b07621aa7802776e Check the logs for full command output. ```

In your code ```python raw_pt = pt[:,:3,:].contiguous() normal = pt[:,3:,:].transpose(1, 2).contiguous() if C > 3 else None ``` Therefore, I was wondering do you used the normal feature except xyz...