RuntimeError during the training example with OIM

[GBJim]

Hi all

After I executed the command python examples/resnet.py -d viper -b 64 -j 2 --loss oim --logs-dir logs/resnet-viper-oim I encountered the following errors:

Process Process-4: Traceback (most recent call last): File "/root/miniconda2/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap self.run() File "/root/miniconda2/lib/python2.7/multiprocessing/process.py", line 114, in run self._target(*self._args, *self._kwargs) File "/root/miniconda2/lib/python2.7/site-packages/torch/utils/data/dataloader.py", line 45, in _worker_loop data_queue.put((idx, samples)) File "/root/miniconda2/lib/python2.7/multiprocessing/queues.py", line 392, in put return send(obj) File "/root/miniconda2/lib/python2.7/site-packages/torch/multiprocessing/queue.py", line 17, in send ForkingPickler(buf, pickle.HIGHEST_PROTOCOL).dump(obj) File "/root/miniconda2/lib/python2.7/pickle.py", line 224, in dump self.save(obj) File "/root/miniconda2/lib/python2.7/pickle.py", line 286, in save f(self, obj) # Call unbound method with explicit self File "/root/miniconda2/lib/python2.7/pickle.py", line 554, in save_tuple save(element) File "/root/miniconda2/lib/python2.7/pickle.py", line 286, in save f(self, obj) # Call unbound method with explicit self File "/root/miniconda2/lib/python2.7/pickle.py", line 606, in save_list self._batch_appends(iter(obj)) File "/root/miniconda2/lib/python2.7/pickle.py", line 639, in _batch_appends save(x) File "/root/miniconda2/lib/python2.7/pickle.py", line 286, in save f(self, obj) # Call unbound method with explicit self File "/root/miniconda2/lib/python2.7/multiprocessing/forking.py", line 67, in dispatcher self.save_reduce(obj=obj, rv) File "/root/miniconda2/lib/python2.7/pickle.py", line 401, in save_reduce save(args) File "/root/miniconda2/lib/python2.7/pickle.py", line 286, in save f(self, obj) # Call unbound method with explicit self File "/root/miniconda2/lib/python2.7/pickle.py", line 554, in save_tuple save(element) File "/root/miniconda2/lib/python2.7/pickle.py", line 286, in save f(self, obj) # Call unbound method with explicit self File "/root/miniconda2/lib/python2.7/multiprocessing/forking.py", line 66, in dispatcher rv = reduce(obj) File "/root/miniconda2/lib/python2.7/site-packages/torch/multiprocessing/reductions.py", line 113, in reduce_storage fd, size = storage._sharefd() RuntimeError: unable to write to file at /py/conda-bld/pytorch_1493669264383/work/torch/lib/TH/THAllocator.c:267

When switch to the xentropy loss with python examples/resnet.py -d viper -b 64 -j 1 --loss xentropy --logs-dir logs/resnet-viper-xentropy The following error occured:

Exception in thread Thread-1: Traceback (most recent call last): File "/root/miniconda2/lib/python2.7/threading.py", line 801, in __bootstrap_inner self.run() File "/root/miniconda2/lib/python2.7/threading.py", line 754, in run self.target(*self.args, *self.__kwargs) File "/root/miniconda2/lib/python2.7/site-packages/torch/utils/data/dataloader.py", line 51, in _pin_memory_loop r = in_queue.get() File "/root/miniconda2/lib/python2.7/multiprocessing/queues.py", line 378, in get return recv() File "/root/miniconda2/lib/python2.7/site-packages/torch/multiprocessing/queue.py", line 22, in recv return pickle.loads(buf) File "/root/miniconda2/lib/python2.7/pickle.py", line 1388, in loads return Unpickler(file).load() File "/root/miniconda2/lib/python2.7/pickle.py", line 864, in load dispatchkey File "/root/miniconda2/lib/python2.7/pickle.py", line 1139, in load_reduce value = func(args) File "/root/miniconda2/lib/python2.7/site-packages/torch/multiprocessing/reductions.py", line 68, in rebuild_storage_fd fd = multiprocessing.reduction.rebuild_handle(df) File "/root/miniconda2/lib/python2.7/multiprocessing/reduction.py", line 155, in rebuild_handle conn = Client(address, authkey=current_process().authkey) File "/root/miniconda2/lib/python2.7/multiprocessing/connection.py", line 169, in Client c = SocketClient(address) File "/root/miniconda2/lib/python2.7/multiprocessing/connection.py", line 308, in SocketClient s.connect(address) File "/root/miniconda2/lib/python2.7/socket.py", line 228, in meth return getattr(self._sock,name)(*args) error: [Errno 111] Connection refused

In both situations, the terminal is frozen after these errors prompt. I have to kill the corresponding Python process in order to exit. Any suggestions to solve this?

[Cysu]

I wonder if it is fine to run the official mnist example?

[GBJim]

Hi @Cysu After going through the MNIST example. No errors happen.

I also tried to train the inception net in example: python examples/inception.py -d viper -b 64 -j 2 --loss xentropy --logs-dir logs/inception-viper-xentropy No errors happen as well.

The interesting thing is that I tried to train ResNet again: The training process froze like the following:, but no errors.

Files already downloaded and verified VIPeR dataset loaded subset | # ids | # images

train | 216 | 432 val | 100 | 200 trainval | 316 | 632 query | 316 | 632 gallery | 316 | 632 Epoch: [0][1/7] Time 160.275 (160.275) Data 0.446 (0.446) Loss 5.375 (5.375) Prec 0.00% (0.00%) Epoch: [0][2/7] Time 0.563 (80.419) Data 0.001 (0.223) Loss 10.057 (7.716) Prec 0.00% (0.00%)

Is this caused by the GPU resource usage? Currently, some Caffe process is also using my GPUs.

[Cysu]

I'm not sure if it is caused by some deadlocks between pytorch and caffe, especially when both are using NCCL. You may try to run it again when the caffe experiments are finished.

[GBJim]

Hi @Cysu Sorry for late response. I tried it again after my Caffe process is terminated.

The training will be frozen when the -j (worker) argument is set to be bigger than 1. If the -j argument is set to be 1, I get error: [Errno 111] Connection refused

[Cysu]

@GBJim Could you please change the num_workers in the official mnist example and see if it has the same problem?

[GBJim]

@Cysu:

I tested the MNIST example with 16 workers. Everything is correct

[Cysu]

Sorry but currently I have no idea why it happened. There should be no much difference between our data loader with the mnist ones. I'm not sure if it is related to using root instead of normal user on Linux.

[GBJim]

Thanks @Cysu I will try to figure it out!

[Cysu]

@GBJim any luck on this?

[GBJim]

Hi @Cysu I've built a new environment for open RE-ID and cloned the latest commit. But it seems like the resnet.py and inception.py are removed from the example folder.

Is there new tutorial of how to do a training or testing? Thanks!

[GBJim]

It seems like the codes are re-organized into oim_loss.py, softmax_loss.py and, triplet_loss.py Let me check if my these scripts can work

[GBJim]

@Cysu

I tried these commands: python examples/oim_loss.py -d viperor python examples/softmax_loss.py -d viper and python examples/triplet_loss.py -d viper as well. Tthe following output is prompted and then the process was frozen. I need to use ctrl+z to exit for the process

root@e50f76502ce4:~/open-reid# python examples/oim_loss.py -d viper
Files already downloaded and verified
VIPeR dataset loaded
  subset   | # ids | # images
  ---------------------------
  train    |   216 |      432
  val      |   100 |      200
  trainval |   316 |      632
  query    |   316 |      632
  gallery  |   316 |      632

[Cysu]

@GBJim Oh, I forgot to update the tutorials. Just finished. Please check here.

Does the previous error still occur when -j 1 is use?

[GBJim]

@Cysu

The process is still frozen when I set to single job. (Maybe I should wait for the process for longer time)

I set job to 1 and tried the following combinations:

OIM + ResNet --> Frozen

OIM + Inception --> RuntimeError: The expanded size of the tensor (128) must match the existing size (64) at non-singleton dimension 1. at /root/pytorch/torch/lib/THC/generic/THCTensor.c:323

SOFTMAX + ResNet --> Frozen

SOFTMAX + Inception --> Works Normally

And thank you for updating the documentation!

[Cysu]

That's weird... What's the script for OIM + Inception?

[GBJim]

@Cysu python examples/oim_loss.py -d viper -a inception -j 1

[lzj322]

I meet the same issue. The problems that @GBJim had happen to me as well. Particularly, this, inception.py has nothing wrong, but resnet.py is Frozen.

[GBJim]

@lzj322 Do you use Nvidia-docker to host the environment?

[lzj322]

@GBJim yes. Would that be a problem? I don't know much about it. I asked the administrator to reset the docker. Now it has normal results. But we don't know why. I am afraid that this issue could happen someday again.

[lzj322]

@GBJim, @Cysu I guess that Dataparallel of pytorch doesn't work well with Nvidia-docker. Or maybe it is caused by pytorch pytorch forum

[Cysu]

@lzj322 Yeah, two programs cannot run on the same device if using NCCL.