Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Question on m2t-motionGPT evaluation metrics #76

Open
OrigamiStationery opened this issue Jan 26, 2024 · 2 comments
Open

Question on m2t-motionGPT evaluation metrics #76

OrigamiStationery opened this issue Jan 26, 2024 · 2 comments

Comments

@OrigamiStationery
Copy link

Thank u for your grand project. It really helps a lot. But now I'm in trouble evaluating the task-specific motionGPT to tackle with motion translation. It seems that when the model is training on m2t task, it still uses the metrics used to evaluate t2m task, which throws an exception below:
lightning_fabric.utilities.exceptions.MisconfigurationException: ModelCheckpoint(monitor='Metrics/FID')could not find the monitored key in the returned metrics: ['epoch', 'step', 'gpt/loss/train', 'total/train', 'Metrics/Matching_score', 'Metrics/gt_Matching_score', 'Metrics/R_precision_top_1', 'Metrics/R_precision_top_2', 'Metrics/R_precision_top_3', 'Metrics/gt_R_precision_top_1', 'Metrics/gt_R_precision_top_2', 'Metrics/gt_R_precision_top_3', 'Metrics/Bleu_1', 'Metrics/Bleu_2', 'Metrics/Bleu_3', 'Metrics/ROUGE_L', 'Metrics/CIDEr', 'Metrics/Bleu_4', 'Metrics/Bert_F1']. HINT: Did you calllog('Metrics/FID', value)in theLightningModule?
I think it might be something wrong in my config file but I don't know how to solve it. desperate for your help OTZ.
btw express my sincere gratitude again.

@billl-jiang
Copy link
Collaborator

Hello,

Thank you for your kind words about the project, and I'm sorry to hear you're encountering issues with evaluating the motionGPT for the motion translation (m2t) task. The problem you're facing with the metrics seems to stem from the task configuration in your model.

For training on the m2t task, it's essential to adjust the validation task to m2t within your configuration file. This can be achieved by setting model.task to m2t in your config file. This adjustment ensures that the appropriate metrics are used for evaluation during the m2t task training.

I hope this helps resolve the issue. Your gratitude is greatly appreciated, and please don't hesitate to reach out if you have further questions.

model:
target: mGPT.models.mgpt.MotionGPT
params:
condition: 'text'
task: 't2m'

@Skye601
Copy link

Skye601 commented Apr 12, 2024

I have set the task to m2t in stage2.yaml, but I still encountered this problem.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants