yqy2001

Hi, `torch.amp.autocast` is used in the [BLIP2-T5 model](https://github.com/salesforce/LAVIS/blob/7aa83e93003dade66f7f7eaba253b10c459b012d/lavis/models/blip2_models/blip2_t5.py#L218), while this API is added in [PyTorch 1.12.0+](https://discuss.pytorch.org/t/module-torch-has-no-attribute-amp/172026/2). An updated requirement for the torch version may be needed, not only to solve...

Different outputs of the official LLaMA repo and transformers' implementation

4

### System Info - `transformers` version: main - Platform: Linux-5.4.0-113-generic-x86_64-with-glibc2.10 - Python version: 3.8.16 - Huggingface_hub version: 0.11.1 - PyTorch version (GPU?): 1.12.1 (True) - Tensorflow version (GPU?): not installed...

Trainer: Functions to inspect model and optimizer status

4

### Feature request In huggingface Trainer, are there any functions to inspect model and optimizer status? such as, how many parameters require grad, learning rate of each parameter, which optimizer...

draft of multimodal openchat training, clip + mlp + mistral (llava arch)

VQGAN training scripts

1

When will the training scripts of VQGAN be released? Thank you in advance.

Thanks for the great work. About training code of VQGAN

2

Thanks for the great work in democratizing MIM for image generation. Are there any plans to release the training of VQGAN? (possibly within the transformers/diffusers/accelerate framework)

Very great, inspiring, and solid work! Thank you!

Thanks for your great work! I noticed that the data used in the May version of LLaVA-NeXT comprises of ~790K data, while the Jan version has only 760K data. Could...