Md. Istiaq Ansari

icon indicating a website link https://istiaqansari.github.io/

icon company Reve Systems Limited icon location Dhaka, Bangladesh icon location Seeing something come to life that I coded is a major source of my dopamine

Results 4 issues of Md. Istiaq Ansari

Parrallel training on Multiple GPU is not included in finetuning process

I would like to train or finetune the models ( in my case SCLSTM ) on multiple GPU using dataparrelle module of pytorch. But it doesn't work.

Word-level padding vs Character-level padding

Hi @helboukkouri The max numbers of letters in a word is set to 50. So for a word with 5 characters is getting padded to 50. For padding, a value...

Training script hangs at torch.Distributed.init_process_group

The training script hangs at [this](https://github.com/helboukkouri/character-bert-pretraining/blob/97160755b685f5970c96c1d6cb1a1c471fef08e9/pretrain_model.py#L420) line. And does nothing after that. I have looked up this suggestion and tried to set word size, Address and Port before that but...

Is there a way to delete or insert certain words/word segments after loading a trained smp model?

I trained a sentencepiece and loaded it to the BPEmb. But there are some words / word segments that I want excluded meaning I want the model to give UNKNOWN...