Jack Cai

I have added code to fintune and evaluate OFA on Scene Graph Generation task. I have used a popular evaluation codebase from the SGG literature to calculate the metrics. Specifically,...

add optional arguments to make sure generate() works without tokenizer being defined

1

# Description Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context. List any dependencies that are required for this change....

[Bug Report] HookedTranformer.generate() with model.tokenizer unset gives pad_token_id error

2

If you are submitting a bug report, please fill in the following details and use the tag [bug]. **Describe the bug** When calling `HookedTranformer.generate()` without initializing the model with a...

Llama Model throwing "RuntimeError: expected scalar type BFloat16 but found Float" when using torch.compile and AMP together

7

### System Info transformers 4.41.0 torch 2.3.0 GPU: NVIDIA GeForce RTX 4090, CUDA version 12.3 ### Who can help? _No response_ ### Information - [ ] The official example scripts...

How to make sure GRPOTrainer is patched when I subclass it

2

In my original script, I subclass the GRPOTrainer for my usecase. I also have an input arg that controls when I want to use UnSloth. I would like to *...