Mengxiangtao

Results 3 comments of Mengxiangtao

@crankyz hava you solved it ?

https://github.com/EleutherAI/knowledge-neurons/blob/be1a84adafd6e973ab0ee012527ac1988632a5e8/knowledge_neurons/knowledge_neurons.py#L197C9-L197C55 ```python baseline_outputs = self.model(**encoded_input) ``` I think this step mainly make gpu memory keep increase, just change it following: ```python with torch.no_grad(): baseline_outputs = self.model(**encoded_input) ``` of course, the...

Thanks, do I just need to modify the configuration file (config.py)?