Facing issue with checkpoint while running mmf_config to Finetune the VisualBERT model for Hateful Meme detection

[AnjumJ123]

If you do not know the root cause of the problem, and wish someone to help you, please post according to this template:

Instructions To Reproduce the Issue:

1. Code for Finetuning the VisualBERT model on Hateful Meme dataset

   
   os.environ['OC_DISABLE_DOT_ACCESS_WARNING']="1"

os.chdir(home)

Define where image features are

feats_dir = os.path.join(home, "features")

Define where train.jsonl is

train_dir = os.path.join(home, "train_v10.jsonl")

!mmf_run config="projects/visual_bert/configs/hateful_memes/from_coco.yaml" \ model="visual_bert" \ dataset=hateful_memes \ run_type=train_val \ checkpoint.max_to_keep=1 \ checkpoint.resume_zoo=visual_bert.pretrained.cc.full \ training.tensorboard=True \ training.checkpoint_interval=50 \ training.evaluation_interval=50 \ training.max_updates=3000 \ training.log_interval=100 \ dataset_config.hateful_memes.max_features=100 \ dataset_config.hateful_memes.annotations.train[0]=$train_dir \ dataset_config.hateful_memes.annotations.val[0]=hateful_memes/defaults/annotations/dev_unseen.jsonl \ dataset_config.hateful_memes.annotations.test[0]=hateful_memes/defaults/annotations/test_unseen.jsonl \ dataset_config.hateful_memes.features.train[0]=$feats_dir \ dataset_config.hateful_memes.features.val[0]=$feats_dir \ dataset_config.hateful_memes.features.test[0]=$feats_dir \ training.lr_ratio=0.3 \ training.use_warmup=True \ training.batch_size=32 \ optimizer.params.lr=5.0e-05 \ env.save_dir=./sub1 \ env.tensorboard_logdir=logs/fit/sub1 \

2. Error Logs:

Namespace(config_override=None, local_rank=None, opts=['config=projects/visual_bert/configs/hateful_memes/from_coco.yaml', 'model=visual_bert', 'dataset=hateful_memes', 'run_type=train_val', 'checkpoint.max_to_keep=1', 'checkpoint.resume_zoo=visual_bert.pretrained.cc.full', 'training.tensorboard=True', 'training.checkpoint_interval=50', 'training.evaluation_interval=50', 'training.max_updates=3000', 'training.log_interval=100', 'dataset_config.hateful_memes.max_features=100', 'dataset_config.hateful_memes.annotations.train[0]=/content/train_v10.jsonl', 'dataset_config.hateful_memes.annotations.val[0]=hateful_memes/defaults/annotations/dev_unseen.jsonl', 'dataset_config.hateful_memes.annotations.test[0]=hateful_memes/defaults/annotations/test_unseen.jsonl', 'dataset_config.hateful_memes.features.train[0]=/content/features', 'dataset_config.hateful_memes.features.val[0]=/content/features', 'dataset_config.hateful_memes.features.test[0]=/content/features', 'training.lr_ratio=0.3', 'training.use_warmup=True', 'training.batch_size=32', 'optimizer.params.lr=5.0e-05', 'env.save_dir=./sub1', 'env.tensorboard_logdir=logs/fit/sub1']) Overriding option config to projects/visual_bert/configs/hateful_memes/from_coco.yaml Overriding option model to visual_bert Overriding option datasets to hateful_memes Overriding option run_type to train_val Traceback (most recent call last): File "/usr/local/bin/mmf_run", line 8, in sys.exit(run()) File "/usr/local/lib/python3.7/dist-packages/mmf_cli/run.py", line 80, in run configuration = Configuration(args) File "/usr/local/lib/python3.7/dist-packages/mmf/utils/configuration.py", line 214, in init self.config = self._merge_with_dotlist(self.config, args.opts) File "/usr/local/lib/python3.7/dist-packages/mmf/utils/configuration.py", line 395, in _merge_with_dotlist " configuration at field {}".format(opt, stripped_field) AttributeError: While updating configuration option checkpoint.max_to_keep is missing from configuration at field max_to_keep

## Expected behavior:

The code should have run as I do not see any obvious errors, but if there is anything which is deprecated as a practice for declaring the config variables with the new release of OmegaConf, need to understand how to mention it in the code.