train-CLIP-FT issues

`train_dataloader` must be implemented to be used with the Lightning Trainer

1

hello I am going to train CLIP with my own dataset, but I met a exception : > raise MisconfigurationException("`train_dataloader` must be implemented to be used with the Lightning Trainer")...

ShuxunoO

resulting checkpoint can be compatible with clip.load()?

1

This repo greatly helped my finetuning process. Thanks. I am wondering if these resulting checkpoints, which are produced by this code in `lightning_logs/version_#/checkpoints/~~~.ckpt,` are compatible with `clip.load`? I could `torch.load`...

KyonP

How to test the model?

2

Thank you for your excellent work. I have trained on my own data set, but I cannot load the trained model. How to test the model?

IvyYZ

`positional_embedding` and `logit_scale` different between `model.model` and `clp`

Is there a reason for this? All other parameters shared by `model.model` and `clp` (in train_finetune.py) are the same. Thank you!

iranroman

Requirements for running the model (Colab) not sufficient

After installing the following packages on Colab: pip install transformers pip install git+https://github.com/openai/CLIP.git pip install 'git+https://github.com/katsura-jp/pytorch-cosine-annealing-with-warmup' pip install pytorch-lightning and running the command: python train_finetune.py --folder data --batch_size 512 --gpu...

bwolfson2