Duke Wang·Nov 23, 2025It’s time to turn off the annoying THINK MODE Qwen-3I really think reasoning is one of the best things LLMs have brought us — but the downside is that it can seriously slow things down…A response icon2A response icon2
Duke Wang·Feb 22, 2025Concise explaination of self.optimizer.zero_grad() in PytorchIn Pytorch you may have seen this function:self.optimizer.zero_grad(). For most of us we already get used to put it at a beginning stage…