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

support training on only LAION 2B #137

Open
anas-awadalla opened this issue Mar 30, 2023 · 5 comments · May be fixed by #261
Open

support training on only LAION 2B #137

anas-awadalla opened this issue Mar 30, 2023 · 5 comments · May be fixed by #261
Labels
enhancement New feature or request good first issue Good for newcomers

Comments

@anas-awadalla
Copy link
Collaborator

          @anas-awadalla Thanks for your quick reply.

Take your running command as an example, how can I change the following command to only train on LAION-2B based on a pre-trained OPT-1.3B?

torchrun --nnodes=1 --nproc_per_node=4 train.py \
--run_name flamingo3B \
--lm_path facebook/opt-1.3b \
--tokenizer_path facebook/opt-1.3b \
--dataset_resampled \
--laion_shards "/path/to/shards/shard-{0000..0999}.tar" \
--mmc4_shards "/path/to/shards/shard-{0000..0999}.tar" \
--batch_size_mmc4 4 \
--batch_size_laion 8 \
--train_num_samples_mmc4 125000 \
--train_num_samples_laion 250000 \
--loss_multiplier_laion 0.2 \
--workers=6 \
--num_epochs 250 \
--lr_scheduler constant \
--warmup_steps 5000 \
--use_media_placement_augmentation \
--mmc4_textsim_threshold 30

By the way, I would like to ask about the contribution of MMC4 for training. Have you conducted an ablation study on MMC4 + LAION-2B and LAION-2B only? Thank you very much for your time and consideration!

Originally posted by @HenryHZY in #129 (comment)

@anas-awadalla
Copy link
Collaborator Author

As implied by the title we should allow users to train on only laion 2B. This should be very straightforward and involves making mmc4 arguments optional in train.py and refactoring the training loop on train_utils.py

@nayeem01
Copy link

nayeem01 commented Aug 8, 2023

@anas-awadalla is it still open for contribution?

@anas-awadalla
Copy link
Collaborator Author

@nayeem01 Yes! We kept the previous PR out for too long and now it is outdated. Feel free to work on it.

@anas-awadalla anas-awadalla linked a pull request Sep 19, 2023 that will close this issue
7 tasks
@isaac-chung
Copy link
Contributor

Noticed the linked PR. Is this open for contribution? @anas-awadalla

@anas-awadalla
Copy link
Collaborator Author

Hi @isaac-chung. We have this implemented in #261. We would still love it if you contributed to a different issue :).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request good first issue Good for newcomers
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants