vince zeng comments

Results 9 comments of


                                            vince zeng

Error when 'make' :warp-ctc/src/ctc_entrypoint.cu(1): error: this declaration has no storage class or type specifier warp-ctc/src/ctc_entrypoint.cu(1): error: expected a ";"

Do u figure out this issue now? If you fix it, could u please tell me how to figure it out?

Error when 'make' :warp-ctc/src/ctc_entrypoint.cu(1): error: this declaration has no storage class or type specifier warp-ctc/src/ctc_entrypoint.cu(1): error: expected a ";"

Do u figure out this issue now? If you fix it, could u please tell me how to figure it out?

About loss

the default param :reduction of ctc loss is 'mean', so why still divide batchsize?

Serious problem in LayoutLMv2: I can't run run_xfun_er.py, as XFUND dataset is not downloadable through the code

hi, have you ever tried downloading the XFUND dataset by yourself and load it from local?

Serious problem in LayoutLMv2: I can't run run_xfun_er.py, as XFUND dataset is not downloadable through the code

so how to load the data form local? can u give me some guidance? (I m new to NLP)

Serious problem in LayoutLMv2: I can't run run_xfun_er.py, as XFUND dataset is not downloadable through the code

`def load_from_disk(dataset_path: str, fs=None, keep_in_memory: Optional[bool] = None) -> Union[Dataset, DatasetDict]` through this interface?

Serious problem in LayoutLMv2: I can't run run_xfun_er.py, as XFUND dataset is not downloadable through the code

thx a lot, I change the code, it work now!

[coati] How to get prompt_path and pretrain_dataset?

> > Is it possible to extract a part of instructions in pretrain_dataset(sft dataset) to generate prompt_datasets? > > I think so, too hihi, 大佬好。想请教您一个问题，我们在训练sft以及prompt的时候，SupervisedDataset()会对instruction进行改写，但是我们在训练reward_model时没有进行改写，在inference.py中也没有进行改写。这样不同训练阶段的输入存在一定差别，不会对模型输出有影响么？

how to use kto_pair loss in the latest version ?

> maybe you can try trl < 0.9.5 fine