Jeffrey Yoo

@ko120 @qiyanjun Looks like the issue is with the `truncate_words_to` keyword argument which isn't part of `GreedyWordSwapWIR`. The argument was add in PR #747. @qiyanjun Could you share the background...

@Yogender11 I'm sorry looks like there was a mistake with the file name. The file `test_logs.json` at https://github.com/QData/TextAttack-A2T/blob/91a20925858801cabeec2b89c598bb146e019b45/evaluate.py#L259 and https://github.com/QData/TextAttack-A2T/blob/91a20925858801cabeec2b89c598bb146e019b45/evaluate.py#L367 both refers to the json log file saved here https://github.com/QData/TextAttack-A2T/blob/91a20925858801cabeec2b89c598bb146e019b45/evaluate.py#L554...

Hi @Han8931. Can you provide more detail as to how you are training your model (e.g. dataset size, attack parameters, etc)?

Thanks! I ended up using `dequantize_4bit` [method ](https://github.com/TimDettmers/bitsandbytes/blob/18e827d666fa2b70a12d539ccedc17aa51b2c97c/bitsandbytes/functional.py#L861)from bnb to dequantize the linear weights, but I think this approach is neat.

Not sure if this is 100% correct way to do it. `dequantize_4bit(module.weight.data, quant_state=module.weight.quant_state)` where module is instance of `bnb.nn.Linear4bit`. That should get you the weight that you can use to...

yes