Xin Lai

icon indicating a website link https://x-lai.github.io/

icon company The Chinese University of Hong Kong icon location HKSAR, China icon location A Ph.D student in the Chinese University of Hong Kong (CUHK)

Results 36 comments of Xin Lai

Availability of ReasonSeg Testing Set

Hi, we are preparing to release an interface (e.g., a website) for the public to submit their testing set results. This work should be done before mid of Oct. Please...

New version <LISA-13B-llama2-v1-explanatory> inference VERY SLOWLY

It works well in my side. Can you give me more detail about this?

AttributeError: 'LISAForCausalLM' object has no attribute 'ce_loss_weight'

Hi, do you want to train further with LISA as the base model?

training on fewer gpus

Can you lower the batch_size and then increase the grad_accumulation_steps to make sure their product keeps the same?

The model "llava-7b-llama-2-7b-chat" merged by myself had problems during training.

I think this is caused by datasets. Can you check whether the datasets are correctly organized?

The number of pred_mask is inconsistent with the number of gt_mask

It seems that `[SEG]` is not added to the vocabulary of the tokenizer successfully. Can you check it by printing out `args.seg_token_idx` after this line (https://github.com/dvlab-research/LISA/blob/main/train_ds.py#L129)? BTW, the issue you...