Frédéric Larochelle @thefredlr

Joined September 2021

Tweets

39
Followers

24
Following

667
Likes

15K

Frédéric Larochelle @thefredlr

4 months ago

@doodlestein @kenwheeler Is rch open source? Sounds interesting

1 0 1 105 0

View Details

Frédéric Larochelle @thefredlr

4 months ago

@devpmxt @sterlingcrispin @martkiro @Polymarket How long does it goes back?

0 0 0 13 0

View Details

Frédéric Larochelle @thefredlr

8 months ago

@embirico I am pretty sure something got broken in the VS Code extension.. I am on Pro, I didn't use it all weekend and suddenly this after a few hours? I just checked and the Codex CLI seems fine.

0 0 1 38 0

View Details

Frédéric Larochelle @thefredlr

a year ago

@predict_addict Small heads up, your link is invalid!

2 0 0 123 0

View Details

Frédéric Larochelle @thefredlr

a year ago

@theo @thegenioo You shoud add the current token count for each conversation, like google AI studios does it. It helps in managing the conversations, not put too much context in.

0 0 1 15 0

View Details

Frédéric Larochelle @thefredlr

a year ago

@theo You should allow user to add and save reusable content blocks (artefacts) that can be quickly reference in chats. Would be great to quickly add a library documentation as context to a conversation. Maybe add a @ mention with a popup to select saved artefacts?

0 0 0 96 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@greenelectrick @KirkLubimov Can already be done with satellites... No need for drones.

0 0 2 44 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@LongBranchMike1 @tcardinalsirois The REM is pretty much the modern version of the line 3 aka the "red line". It was a metro line that was in the original plans for the metro and it was supposed to use the Mont-Royal tunnel. Then Montreal won the bid for the 1976 Olympics and they focused on the green line.

0 0 6 189 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@KittiesRules @OliParle Vanier

0 0 0 127 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@jimnasyum @robkhenderson She is 20...

1 0 1 296 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@MitziMasha @sivori Some field like medicine have way higher divorce rates, medecine, law, ... Others like engineering, it's less than 10%.

0 0 1 194 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@BK_McCallum @peterrhague One Starship will already be slighty bigger than the volume of the ISS...

0 0 1 30 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@UltraLiberalKid @RM_Transit You forgot the most critical extension. The orange line to Bois-Franc!

0 0 4 55 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@karpathy @cHHillee I would be curious to see where JAX is in this little competition...

0 0 3 2K 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@mov_axbx @__tinygrad__ @EricHallahan Found this: github.com/intel/compute-… :( The question now is the hardware support it (and is simply disabled) or not.

1 0 0 98 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@EricHallahan @__tinygrad__ @mov_axbx Yep :(

0 0 1 153 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@__tinygrad__ @mov_axbx @EricHallahan From what I could find, they have a P2P benchmark on their compute-benchmarks repo here: github.com/intel/compute-… It is only supported for Level Zero, not OpenCL. There is also a thread from FluidX3D from last May on P2P support with OpenCL and it was broken on all platforms.

2 0 1 290 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@__tinygrad__ *reading somewhere

0 0 0 71 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@__tinygrad__ I remember somewhere that Nvidia was planning to push the release cycle from 24 months to 30 months. Soo probably more than one year away from coming out?

1 0 1 548 0

View Details

Frédéric Larochelle @thefredlr

2 years ago

@fabinulleins @KikeljGregor @__tinygrad__ @intel @intelGaming @intelGraphics Oh wait, I might be wrong, it quoted the GDDR6 speed per chip on the A770. But anyways, no GPUs is even really close to filling PCIe gen 4x16...