Unable to train model with paris lille dataset

[Sajjadmanal]

Hi,
I tried training the model with paris lille dataset but get the below error:

Epoch 0/99

Traceback (most recent call last):
File "C:/Users/sm50014/Desktop/ms_deepvoxscene/apps/train.py", line 172, in
best_val_model, best_train_model, model = trainer.train_model()
File "C:\Users\sm50014\Desktop\ms_deepvoxscene\utils\trainer.py", line 122, in train_model
self.train_one_epoch(epoch)
File "C:\Users\sm50014\Desktop\ms_deepvoxscene\utils\trainer.py", line 177, in train_one_epoch
for i,data_batch in enumerate(self.dset_loaders[phase]):
File "C:\Users\sm50014\AppData\Local\Continuum\anaconda3\envs\lidar\lib\site-packages\torch\utils\data\dataloader.py", line 193, in iter
return _DataLoaderIter(self)
File "C:\Users\sm50014\AppData\Local\Continuum\anaconda3\envs\lidar\lib\site-packages\torch\utils\data\dataloader.py", line 469, in init
w.start()
File "C:\Users\sm50014\AppData\Local\Continuum\anaconda3\envs\lidar\lib\multiprocessing\process.py", line 105, in start
self._popen = self._Popen(self)
File "C:\Users\sm50014\AppData\Local\Continuum\anaconda3\envs\lidar\lib\multiprocessing\context.py", line 223, in _Popen
return _default_context.get_context().Process._Popen(process_obj)
File "C:\Users\sm50014\AppData\Local\Continuum\anaconda3\envs\lidar\lib\multiprocessing\context.py", line 322, in _Popen
return Popen(process_obj)
File "C:\Users\sm50014\AppData\Local\Continuum\anaconda3\envs\lidar\lib\multiprocessing\popen_spawn_win32.py", line 65, in init
reduction.dump(process_obj, to_child)
File "C:\Users\sm50014\AppData\Local\Continuum\anaconda3\envs\lidar\lib\multiprocessing\reduction.py", line 60, in dump
ForkingPickler(file, protocol).dump(obj)
OSError: [Errno 22] Invalid argument

I am using below configuration:

###############################################################################
DATASET: "parislille3d"
#DATASET: "semantic3d"
#DATASET: "s3dis"
    
###############################################################################
NB_SAMPLES: 10 # Number of training with different initialization of weights

###############################################################################
#PHASE_LIST: ['train'] # Use all point clouds to train the model
PHASE_LIST: ['val','train'] # use alternatively each point cloud of the dataset as validation of a model trained on all others point clouds
    
###############################################################################
NUM_WORKERS: 8 # Number of threads for pytorch Dataloader
DEVICE_ID: 0 # GPU id
    
###############################################################################
### use this to use (or not) the color or reflectance channel in input of your networks
#USE_COLOR: True
#USE_COLOR: False
#USE_REFLECTANCE: True
#USE_REFLECTANCE: False
    
###############################################################################
#SCALES: {1}
SCALES: {1,2}
#SCALES: {1,2,3}
### size of occupancy grid (GRID_SIZE x GRID_SIZE x GRID_SIZE), GRID_SIZE should be <= 64
#GRID_SIZE: 16
GRID_SIZE: 32

###############################################################################
### size of each voxel of occupancy grid (VOXEL_SIZE m x VOXEL_SIZE m x VOXEL_SIZE m)
VOXEL_SIZE: 0.05
#VOXEL_SIZE: 0.10
#VOXEL_SIZE: 0.15
#VOXEL_SIZE: 0.20
#VOXEL_SIZE: 0.40

###############################################################################    
BATCH_SIZE: 20
NB_POINTS_PER_CLASS: 1000 # number of points randomly chosen in each class at beginning of each epoch
NB_POINTS_PER_CLASS_VAL: 1000
NUM_EPOCHS: 100

###############################################################################
### intialization method for weights of the network
#INITIALIZATION: "xavier"
INITIALIZATION: "kaiming"
#INITIALIZATION: "orthogonal"

###############################################################################
OPTIMIZER: "Adam"
#OPTIMIZER: "Adadelta"
#OPTIMIZER: "RMSprop"
#OPTIMIZER: "SGD"

###############################################################################
EPS: !!float 1e-8 # epsilon parameter for optimizers
    
###############################################################################
LR: !!float 0.001 # Learning Rate
    
###############################################################################
WD: !!float 0 # Weight Decay
    
###############################################################################
#GRADIENT_CLIPPING: True
GRADIENT_CLIPPING: False

###############################################################################
### pytorch learning rate scheduler among the following: [None, "ExponentialLR", "CosineAnnealingLR"]
### except for None, lr_scheduler parameters are set such as at last epoch, lr = 1e-6
#LR_SCHEDULER: None
LR_SCHEDULER: "ExponentialLR"
#LR_SCHEDULER: "CosineAnnealingLR"
  
###############################################################################
### Model name among the following: ["VoxNet", "ModVoxNet", "VGGLikeNet", "ModVGGLikeNet"]
MODEL_NAME: "VoxNet"
#MODEL_NAME: "ModVoxNet"
#MODEL_NAME: "VGGLikeNet"
#MODEL_NAME: "ModVGGLikeNet"

How to train paris lille dataset and using which model for best accuracy?