Training from scrach

[rginjapan]

Training from scratch. Set "--resume" (along with "--run_id" and "--step") to resume from a checkpoint.
Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/3
Training from scratch. Set "--resume" (along with "--run_id" and "--step") to resume from a checkpoint.
Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/3

distributed_backend=nccl
All distributed processes registered. Starting with 3 processes

Logging everything in /media/sly2/860EVO/spoc-robot-training/log/YBqLm86o
LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2]
LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2]
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2]

| Name | Type | Params

0 | model | EarlyFusionCnnTransformer | 226 M
1 | frames_metric | SumMetric | 0

226 M Trainable params
0 Non-trainable params
226 M Total params
906.443 Total estimated model params size (MB)
Sanity Checking DataLoader 0: 0%|

Then the error occurs:
File "/media/spoc-robot-training/online_evaluation/local_logging_utils.py", line 66, in version
return self._experiment.id
AttributeError: 'NoneType' object has no attribute 'id'

This is the command I used:
python -m training.offline.train_pl --dataset_version /media/spoc-robot-training/training_data/fifteen/ObjectNavType --eval_every 50 --precision 16-mixed --input_sensors raw_navigation_camera raw_manipulation_camera last_actions an_object_is_in_hand --per_gpu_batch 4 --sliding_window 50 --model EarlyFusionCnnTransformer --model_version siglip_base_3 --lr 0.0002 --wandb_logging False --output_dir /media/spoc-robot-training/log