Matteo Grella
@marco-nicola
Hi @nikolaydubina, Yes, that's on our TODO list. By the way, if you had to choose, would you prefer benchmarks on single operations (e.g., matrix-vector multiplication) or on higher level...
It's decided then. To start, we will make a direct comparison of the basic operators of the spaGO auto-grad package with the PyTorch ones. Do you feel like helping on...
Hey @jimidle, Thanks for this! I'm wondering if you have time to support us on this topic, using Antora and ASCII doc as you suggested. Just in case, in [this...
Thanks @Tonghua-Li to experiment spaGO on Chinese models! Let me take a look. In the meantime, did you check already if the output from the tokenization matches the one in...
Solved #101
I actually think there's a lot of room for improvement on that front! However, I wonder how to recover an error in case of a size mismatch, for example. If...
Hello both of you @jimidle and @abishekmuthian. I apologize for taking so long to respond! I somehow missed the GitHub notifications, and @marco-nicola was sure I had already dealt with...
Hey @abishekmuthian, It was a simple _True_ flag that was supposed to be a _False_ flag that confused us both! Hence, with spaGO, output is forced to be a distribution...