WANDB_MODE=offline accelerate launch --num_machines=1 --num_processes=8 --machine_rank=0 --main_process_ip=127.0.0.1 --main_process_port=9999 --same_network scripts/train_titok.py config=configs/training/TiTok/stage1/titok_b64.yaml
experiment.project="titok_b64_stage1"
experiment.name="titok_b64_stage1_run1"
experiment.output_dir="titok_b64_stage1_run1"
training.per_gpu_batch_size=32

UnboundLocalError: local variable 'train_dataloader' referenced before assignment
1.When I used the stage1 command line, I found that the dataloader was not defined in the config.How can I load this
2.Since I haven't run the whole experiment yet, I intercepted the eval code and directly input two images into the eval of titok. However, the rFID and IS obtained are not ideal. The rFID is 24 and the IS is 1.9, which is quite different from the results given in the paper. Have you encountered such a problem?

WANDB_MODE=offline accelerate launch --num_machines=1 --num_processes=8 --machine_rank=0 --main_process_ip=127.0.0.1 --main_process_port=9999 --same_network scripts/train_titok.py config=configs/training/TiTok/stage1/titok_b64.yaml
experiment.project="titok_b64_stage1"
experiment.name="titok_b64_stage1_run1"
experiment.output_dir="titok_b64_stage1_run1"
training.per_gpu_batch_size=32
1.When I used the stage1 command line, I found that the dataloader was not defined in the config.How can I load this
2.Since I haven't run the whole experiment yet, I intercepted the eval code and directly input two images into the eval of titok. However, the rFID and IS obtained are not ideal. The rFID is 24 and the IS is 1.9, which is quite different from the results given in the paper. Have you encountered such a problem?