Spaces:

Salesforce

/

BLIP2

like 588

Runtime error

Finetuning examples can be found in https://github.com/salesforce/LAVIS/tree/main/lavis/projects/blip2/train.

Support for colab finetuning will most likely not happening. The model won't fit the VRAM for training with a reasonable batch size.

Sorry, this is extremely confusing and useless

@dxli could you possibly provide the total amount of VRAM necessary to attempt a fine-tune?
Thank you very much for your time.

any update?

Do you have a huggingface - pytorch implementation of this ?

@dxli

Could we please get some proper guidance on fine tuning this model? There are many use cases for it. Is training it possible with the HuggingFace Trainer for example?
The provided finetuning examples are not helpful.