<jonpsy[m]>
<EshaanAgarwal[m]> " would it be possible to..." <- Night, perhaps?
<EshaanAgarwal[m]>
<jonpsy[m]> "Night, perhaps?" <- Sure ! What time ?
<fieryblade[m]>
<EshaanAgarwal[m]> "Sure ! What time ?..." <- Anytime after 8 works fine for me, but wait for jonpsy to answer before re-scheduling.
<jonpsy[m]>
Same, 8
<EshaanAgarwal[m]>
<jonpsy[m]> "Same, 8" <- I might have some difficulties with night. Perhaps can we do it tomorrow ? 🥲
<fieryblade[m]>
<EshaanAgarwal[m]> "I might have some difficulties..." <- I need to connect with you for a maximum of 20 mins, so provide me with the earliest time tomorrow you can be available for a meet. Also, if jonpsy is not able to attend at that time, maybe we can create two separate meetings. This will prevent any further delay, I think.
<EshaanAgarwal[m]>
<fieryblade[m]> "I need to connect with you for a..." <- In that case, would you be available anytime after 11 AM IST ? I can schedule a separate meet with jonpsy:
<fieryblade[m]>
<EshaanAgarwal[m]> "In that case, would you be..." <- yes, I am available
<jonpsy[m]>
Me too
<EshaanAgarwal[m]>
<fieryblade[m]> "yes, I am available" <- I will schedule a meet at 12 AM IST then.
<jonpsy[m]>
perhaps you mean PM?* ;)
<EshaanAgarwal[m]>
* I will schedule a meet at 12 PM IST then.
<rcurtin[m]>
the "short story" is that when serialization is handled correctly, neural networks can take under 1GB of RAM to compile and can be 4x as fast to compile
<rcurtin[m]>
I think there could still be ways to reduce that further, but, maybe that can happen some other day :)