I'd like to write a blog post about this (I write tech articles, this seems very cool subject) I know how RoBERTa works and I've used it many times for the experiments I write about. Do you have any suggestions what dataset can I use since you did this y-combinator start-ups. I don't want to copy your work.
Sounds awesome! No worries with this, maybe ScrapingBee or something like that could help you fill out the dataset, the Crunchbase API might also be good for this
@@connor-shorten good. I'll check it out, and see if I can replicate the procedure. i have been using just beautiful soup to collect various data from the sites. also i used RoBERTa mostly for classifications purposes. So, no need for tuning? I use the top layer as is, right? (sorry to bother im excited on this one)
@@iva1389 I haven’t yet experimented with ranking so I can’t advise from concrete experience unfortunately, but yeah it sounds reasonable to leave the embeddings frozen and just fine tune the ranking heads (like maybe 2-3 layers of MLP should be all you need I hope)
this is awesome example of practical vector search. please do a video with importing dataset to a base, as you suggested since we're aren't all from a coding background but would like to know more. thanks
Hello there, I'm working on my early stage startup xariia, whose vision is to automate custom fashion by generating custom print designs for cloths based on user preferences. I came across your content while researching GANs, and I was wondering if you'd be interested in becoming a partner?