Runnin C4 W3 Lab 3: Question Answering with BERT and HuggingFace on local PC / Mac

Course Q&A Natural Language Processing NLP with Attention Models
1

Has any of you tried running this notebook on your local pc or mac?

Especially for the cell # 20, while it took about 15 minutes on Coursera server, it took 5 hours on Mac. I couldn’t run on a Windows PC as Google Trax doesn’t install on it.

In the light of this, will a Mac Pro with M1 Pro / Max chipset run this piece faster? What’s your experience?

2

Hi Shantimohan_Elchuri,

I don’t know if this is useful to you, but here is my experience.

I ran the Question Answering with BERT and HuggingFace on my Ubuntu system with an NVIDIA 3090, and it took 1 minute and 35 seconds to finish training.

In the past I tried training NNs on my 2017 MacBook Pro with a Radeon Pro 560, and although I managed to get it to work at times, I could not make it work reliably. Because of this, I have been waiting for convincing demonstrations on how to use the M1 reliably to train large NNs.