TRAIN.FINETUNE: command not found`

[Joan-gao]

hi, I followed install instruction, but in eval phase I have follow error:
Traceback (most recent call last): File "run_net_emamix.py", line 44, in <module> main() File "run_net_emamix.py", line 23, in main cfg = load_config(args) File "D:\files\AndroidStudioProjects\SVFormer\timesformer\utils\parser.py", line 77, in load_config cfg.merge_from_file(args.cfg_file) File "D:\program_files\anaconda\envs\svformer\lib\site-packages\fvcore\common\config.py", line 121, in merge_from_file self.merge_from_other_cfg(loaded_cfg) File "D:\program_files\anaconda\envs\svformer\lib\site-packages\fvcore\common\config.py", line 132, in merge_from_other_cfg return super().merge_from_other_cfg(cfg_other) File "D:\program_files\anaconda\envs\svformer\lib\site-packages\yacs\config.py", line 217, in merge_from_other_cfg _merge_a_into_b(cfg_other, self, self, []) File "D:\program_files\anaconda\envs\svformer\lib\site-packages\yacs\config.py", line 478, in _merge_a_into_b _merge_a_into_b(v, b[k], root, key_list + [k]) File "D:\program_files\anaconda\envs\svformer\lib\site-packages\yacs\config.py", line 491, in _merge_a_into_b raise KeyError("Non-existent config key: {}".format(full_key)) KeyError: 'Non-existent config key: TRAIN.GAMMA' train.sh: line 10: TRAIN.FINETUNE: command not found
I wonder how to deal with this problem?

[XinyuSun]

just add a \ after the line:9 TRAIN.ENABLE False in train.sh

python setup.py build develop
python run_net_emamix.py \
  --cfg configs/Kinetics/TimeSformer_base_ssl.yaml \
  DATA.PATH_TO_DATA_DIR ./dataset/list_k400_1/ \
  OUTPUT_DIR ./output/list_k400_1/k400_1/ \
  NUM_GPUS 8 \
  TRAIN.BATCH_SIZE 8 \
  TEST.BATCH_SIZE 64 \
  TRAIN.ENABLE False \
  TRAIN.FINETUNE False \
  TRAIN.CHECKPOINT_FILE_PATH ./checkpoints/checkpoint.pyth

[RuijieSpace]

just add a \ after the line:9 TRAIN.ENABLE False in train.sh

python setup.py build develop
python run_net_emamix.py \
  --cfg configs/Kinetics/TimeSformer_base_ssl.yaml \
  DATA.PATH_TO_DATA_DIR ./dataset/list_k400_1/ \
  OUTPUT_DIR ./output/list_k400_1/k400_1/ \
  NUM_GPUS 8 \
  TRAIN.BATCH_SIZE 8 \
  TEST.BATCH_SIZE 64 \
  TRAIN.ENABLE False \
  TRAIN.FINETUNE False \
  TRAIN.CHECKPOINT_FILE_PATH ./checkpoints/checkpoint.pyth

Hello, have you run through this code? Is the code provided by the author complete? I have encountered some problems. Can you leave an email to communicate? Thank you very much.

[XinyuSun]

just add a \ after the line:9 TRAIN.ENABLE False in train.sh

python setup.py build develop
python run_net_emamix.py \
  --cfg configs/Kinetics/TimeSformer_base_ssl.yaml \
  DATA.PATH_TO_DATA_DIR ./dataset/list_k400_1/ \
  OUTPUT_DIR ./output/list_k400_1/k400_1/ \
  NUM_GPUS 8 \
  TRAIN.BATCH_SIZE 8 \
  TEST.BATCH_SIZE 64 \
  TRAIN.ENABLE False \
  TRAIN.FINETUNE False \
  TRAIN.CHECKPOINT_FILE_PATH ./checkpoints/checkpoint.pyth

Hello, have you run through this code? Is the code provided by the author complete? I have encountered some problems. Can you leave an email to communicate? Thank you very much.

Feel free to contact me using wechat Anthracenesy

[t-rizvi]

just add a \ after the line:9 TRAIN.ENABLE False in train.sh

python setup.py build develop
python run_net_emamix.py \
  --cfg configs/Kinetics/TimeSformer_base_ssl.yaml \
  DATA.PATH_TO_DATA_DIR ./dataset/list_k400_1/ \
  OUTPUT_DIR ./output/list_k400_1/k400_1/ \
  NUM_GPUS 8 \
  TRAIN.BATCH_SIZE 8 \
  TEST.BATCH_SIZE 64 \
  TRAIN.ENABLE False \
  TRAIN.FINETUNE False \
  TRAIN.CHECKPOINT_FILE_PATH ./checkpoints/checkpoint.pyth

Hello, have you run through this code? Is the code provided by the author complete? I have encountered some problems. Can you leave an email to communicate? Thank you very much.

Feel free to contact me using wechat Anthracenesy

Can you help to run this repo, I have posted it as new issue

[XinyuSun]

maybe you can try this forked repo:
https://github.com/XinyuSun/SVFormer

[t-rizvi]

Sure, i will check and let you know

[t-rizvi]

maybe you can try this forked repo:
https://github.com/XinyuSun/SVFormer

I trained a pretrained network from internet that seems incorrect as train error values are approaching zero. Can you share the pre-trained model used in this repository ?