athn-nik / sinc Goto Github PK

Thanks for your great work!
However, when I run your code follow the instruction, I meet the error below. It seems to be a problem related to the config file. How can I solve the problem？

omegaconf.errors.InterpolationKeyError: Interpolation key 'callback.render.nvids_to_save' not found

Hi, thank you for releasing the great work! I am a little confused on what data the model is being trained on in the provided code and instruction. Looking at the instruction, it appears that the model is only trained on real data and not on the synthetic GPT-3 guided compositional motion data. Would it be possible for you to provide the model trained on both real and synthetic data? If not, would it be possible for you to provide the synthetic data and instructions on how I can train the model on both real and synthetic data shown in the paper. Out of curiosity, how long does training take? Thank you!

Hey! It looks like your private api key has been uploaded to git, I recommend to remove it :)

Hi,
I'm tryng to run Step 2 (b) but I don't understand where i can find the file ".../latest-epoch=599.ckpt" required to initiate the state_dict.
Can you explain how I can obtain it?

Exception has occurred: FileNotFoundError
[Errno 2] No such file or directory: '/is/cluster/fast/nathanasiou/logs/sinc/sinc-arxiv/temos-bs64x1-scheduler/babel-amass/checkpoints/latest-epoch=599.ckpt'
File "/andromeda/personal/lmandelli/sinc/train.py", line 121, in train
state_dict = torch.load('/is/cluster/fast/nathanasiou/logs/sinc/sinc-arxiv/temos-bs64x1-scheduler/babel-amass/checkpoints/latest-epoch=599.ckpt', map_location='cpu')
File "/andromeda/personal/lmandelli/sinc/train.py", line 50, in _train
return train(cfg, ckpt_ft)
File "/andromeda/personal/lmandelli/sinc/train.py", line 195, in
_train()
FileNotFoundError: [Errno 2] No such file or directory: '/is/cluster/fast/nathanasiou/logs/sinc/sinc-arxiv/temos-bs64x1-scheduler/babel-amass/checkpoints/latest-epoch=599.ckpt'

Error occurs when sampling:
Error locating target 'teach.data.BABELDataModule', set env var HYDRA_FULL_ERROR=1 to see chained exception.
full_key: data
AttributeError: module 'teach.data' has no attribute 'BABELDataModule'

During handling of the above exception, another exception occurred:

ModuleNotFoundError: No module named 'teach.data.BABELDataModule'

where should I put the teach folder and how to fix this issue?

Hello, thank you very much for providing such an excellent project.
I noticed that you made a comparison with TEMOS in paper. I have trained TEMOS before, but it is implemented on the KIT-ML dataset, so I would like to ask:

1. How to train TEMOS on the BABEL dataset？
2. How to train SINC on the KIT-ML dataset？

Looking forward to your reply, thank you！

Thanks for your nice work! Could you provide the best pretrained model in the paper?