the same happens with BloomZ, and that is listed as open
the same happens with BloomZ, and that is listed as open
neural network weights are just files, collections of numbers forming matrices; how is a partially open collection of weights of any use
the weights are open
$ docker exec -it ollama ollama show gemma:7b
Model
arch gemma
parameters 9B
quantization Q4_0
context length 8192
embedding length 3072
Parameters
stop "<start_of_turn>"
stop "<end_of_turn>"
penalize_newline false
repeat_penalty 1
License
Gemma Terms of Use
Last modified: February 21, 2024
how are the weights partially open?
130,000 units, I don’t know how
grabs popcorn
I am using this https://github.com/lucaspar/firefox-vertical-tabs/ ; same idea
it needs to collapse automatically, otherwise it just wastes a lot of space
Well, what’s the problem. They have bacon and they have ice cr… oh I see the error now. Just add a generic response the ice cream machine is broken and move on!
These things can just be unique fields. I think the takeaway here is exactly to not use these unique fields as database keys if you have the option / if it’s up to you.
I received an email about this one year ago from one of the universities technicians/systems maintainer to let them know if we were running non-openjdk java runtimes because they received a million-dollar “warning” email. Greedy corps are even going after universities.
if your tractor can’t run farming simulator, is it even a tractor?
surely these are things that should be considered, but they move in relation to what? And is this surprising amount of any significance for tens or hundreds of miles of rail?
yeah, who knows - some people might rather waste time fighting a language model than actually learning things…
it’s too early for something like this imo, but maybe in a few years it serves a purpose
you’re thinking anywhere on the platform, I’m suggesting a known place near a station by which the train passes and its location - at that moment - is known.
All the system needs is a ground-truth location after a certain amount of time. GPS is just a cheap and convenient way to do it almost anywhere, but this location correction doesn’t need to be satellite-based at all.
the initial location doesn’t need to be GPS, just a known anchor location. Which is trivial to implement in the case of trains, since stations don’t move that drastically.
a lot cooler
ice what you did there
long-term inertial guidance
this is exactly what I got from the article: a more accurate inertial navigation. What part violates relativity?
“replacing GPS” is a stretch, but it’s some sci-fi tech to use when GPS can’t be used
Version 2024.6.2 does not include the final fix for this issue. We are still completing our internal testing, but we plan to submit another release to the store later this week as soon as it is verified.
from the github issue
Finally! And it seems it’s the 2024.6.2 with the actual fix (2024.6.0 seems to have promised a fix that didn’t cover all cases),
tl;dr: AI + competition