clickwiseorange

5 ms·

1. 1.▲
   by clickwiseorange 2y ago

   1. The Nature paper said one thing, the code did something else, as we've discovered. The RL method does some training as it goes. So, pre-training is not the same as training. Hence "pre". Another problem with pretraining in
2. 2.▲
   by clickwiseorange 2y ago

   The odd thing is that they don't compute timing in RL, but claim that somehow TNS and WNS improved. Does anyone believe this? With five circuits and three wins, the results are a coin toss.
3. 3.▲
   by clickwiseorange 2y ago

   It's actually not clear who was bullied. The two researchers ganged up on Chatterjee and got him fired because he used the word "fraud" - wrongful termination of a whistleblower. Only recently Google settled with Chatterjee f
4. 4.▲
   by clickwiseorange 2y ago

   Good question. It's not just ibm14, but everything people outside Google tried shows that RL is much worse than prior methods. NVDLA, BlackParrot, etc. There is a strong possibility that Google pre-trained RL on certain TPU designs the
5. 5.▲
   by clickwiseorange 2y ago

   It's not that one needs an excuse. The Google CT repo said clearly you don't need to pretrain. "supported" usually includes at least an illustration, some scripts to get it going - no such thing there before Kahng's
6. 6.▲
   by clickwiseorange 2y ago

   Oh, man... this is the same old stuff from the 2023 Anna Goldie statement (is this Anna Goldie's comment?). This was all addressed by Kahng in 2023 - no valid criticisms. Where do I start? Kahng's ISPD 2023 paper is not in dispute