Ttbd

Results 4 comments of Ttbd

@thequilo Hey!the same problem My config is as follow: ``` def _loss_names(d): ret = { "itm": 0, "mlm": 0, "mpp": 0, "vqa": 0, "nlvr2": 0, "irtr": 0, } ret.update(d) return...

> ImportError: cannot import name 'convert_norm_act' from 'timm.models.layers' Do you have such a question I downloaded the files from https://github.com/huggingface/pytorch-image-models/ and Replace the entire timm folder. It works for me

我设置了这两个参数之后 NCCL_IB_DISABLE=1 NCCL_P2P_DISABLE=1 deepspeed train.py 训练代码的grad_norm一直都是0.0,但是loss有在降低,这样正常吗?

Hi @sorobedio and @FantasticZihao, I'm currently trying to work with the VQGAN OpenImages (f=8), 256 model, but I'm unable to locate the corresponding config.yaml file. Could you guys please share...