The performance of RandLA-Net trained on Semantic3D

### Checklist

- [X] I have searched for [similar issues](https://github.com/isl-org/Open3D-ML/issues).
- [X] I have tested with the [latest development wheel](http://www.open3d.org/docs/latest/getting_started.html#development-version-pip).
- [X] I have checked the [release documentation](http://www.open3d.org/docs/release/) and the [latest documentation](http://www.open3d.org/docs/latest/) (for `master` branch).

### My Question

Thanks for your excellent work! I have a problem of reproducing the official results of RandLA-Net trained on Semantic3D. I trained RandLA-Net on Semantic3d with official config file (ml3d/configs/randlanet_semantic3d.yml), but the result I got has a large gap below the official result. The following is the final training result and is this result reasonable for training? And which split is the official reported performance from, train, val or test? Looking forward for your reply.
```yml
dataset:
name: Semantic3D
dataset_path: # path/to/your/dataset
cache_dir: ./logs/cache_small3d/
class_weights: [5181602, 5012952, 6830086, 1311528, 10476365, 946982, 334860, 269353]
ignored_label_inds: [0]
num_points: 65536
test_result_folder: ./test
use_cache: true
val_files:
- bildstein_station1_xyz_intensity_rgb
- domfountain_station1_xyz_intensity_rgb
steps_per_epoch_train: 500
steps_per_epoch_valid: 10
model:
name: RandLANet
batcher: DefaultBatcher
ckpt_path: I set this as the weight file download from model_zoo.md
num_neighbors: 16
num_layers: 5
num_points: 65536
num_classes: 8
ignored_label_inds: [0]
sub_sampling_ratio: [4, 4, 4, 4, 2]
in_channels: 6
dim_features: 8
dim_output: [16, 64, 128, 256, 512]
grid_size: 0.06
augment:
recenter:
dim: [0, 1]
normalize:
feat:
method: linear
bias: 0
scale: 255
rotate:
method: vertical
scale:
min_s: 0.9
max_s: 1.1
noise:
noise_std: 0.001
pipeline:
name: SemanticSegmentation
num_workers: 0
pin_memory: False
optimizer:
lr: 0.001
batch_size: 2
main_log_dir: ./logs
max_epoch: 100
save_ckpt_freq: 5
scheduler_gamma: 0.9886
test_batch_size: 1
train_sum_dir: train_log
val_batch_size: 2
summary:
record_for: []
max_pts:
use_reference: false
max_outputs: 1

```