search

facebookresearch / mmf

A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)
https://mmf.sh/
Other
5.48k stars 935 forks source link

Issue in training a model using Hateful Memes Data set #509

Closed AjinkyaP1998 closed 4 years ago

AjinkyaP1998 commented 4 years ago

❓ Questions and Help

For the Hateful Memes challenge, after downloading the data and converting it to the MMF format. I executed the following code for training a model:

!mmf_run config=projects/hateful_memes/configs/mmbt/defaults.yaml \ model=mmbt \ dataset=hateful_memes \ training.log_interval=50 \ training.max_updates=3000 \ training.batch_size=16 \ training.evaluation_interval=500

And in between the execution of the code, some space opened up for the user's input. What is this input for? What am I suppose to put in it to train the model?

AjinkyaP1998 commented 4 years ago

After some time of execution of the above-mentioned code, an error pops up saying:

RuntimeError: DataLoader worker (pid 666) is killed by signal: Killed.

gireek commented 4 years ago

I am not really sure but my issue got solved when I reduced training batch size. Do try that.

vedanuj commented 4 years ago

Can you provide a more detailed stack trace? Also please attach environment details for bug reports.

apsdehal commented 4 years ago

Closing this as there hasn't been a response and I believe recommendation to decrease batch size solves this. If the issue persists, please open up a new one with details.