Mat Kelcey
@matpalm.com
📤 405
📥 129
📝 132
data nerd wannabe
https://matpalm.com/
reposted by
Mat Kelcey
Edge Impulse
10 months ago
At Imagine Innovators, we just announced our Developer Plan, offering expanded access to Edge Impulse’s powerful tools, completely free! Get GPU training, longer runtimes, more projects, and production-ready licensing, all at zero cost. Details:
www.edgeimpulse.com/blog/introdu...
loading . . .
Introducing the Developer Plan: More Power, More Freedom — for Free
With our new Developer pricing plan you’ll now get the best of our community and professional plans. That means GPU training, longer runtimes, more projects, and production-ready licensing, all at zer...
https://www.edgeimpulse.com/blog/introducing-the-developer-plan/
0
4
3
have moved mixing out of the rack to the model1. such a great performance focussed mixer!
#improv
#techno
loading . . .
10 months ago
0
4
0
reposted by
Mat Kelcey
Flux
11 months ago
It's about time I updated my recommended
#FPGA
links. Send me a suggestion this
#FPGAFriday
. 😊
projectf.io/recommended-...
loading . . .
Recommended FPGA Sites
This page collects some of my favourite FPGA resources from across the Internet. FPGA Learning Resources Black Mesa Labs FPGA Tutorial Bruno Levy’s Learn FPGA fpga4fun.com fpgacpu.ca Nandland Project ...
https://projectf.io/recommended-fpga-sites/
0
10
2
i wish this was always an option...
12 months ago
0
3
0
still "finding" ( i.e. stepping on ) screws from the great M3 drawer spill of Sept '24
12 months ago
0
0
0
reposted by
Mat Kelcey
jake
about 1 year ago
I'll take "n^2 - n cursed images" for $1000, Alec
add a skeleton here at some point
1
6
3
reposted by
Mat Kelcey
Sander Dieleman
about 1 year ago
Great interview with
@jascha.sohldickstein.com
about diffusion models! This is the first in a series: similar interviews with Yang Song and yours truly will follow soon. (One of these is not like the others -- both of them basically invented the field, and I occasionally write a blog post 🥲)
loading . . .
History of Diffusion - Jascha Sohl-Dickstein
YouTube video by Bain Capital Ventures
https://www.youtube.com/watch?v=VpYNlHIHT7o
0
43
11
reposted by
Mat Kelcey
about 1 year ago
Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n
3
95
36
"In what part of the world do you currently live?" lol
about 1 year ago
0
3
0
keep coming back to "Scalable MatMul-free Language Modeling"
arxiv.org/abs/2406.02528
full of fun ideas! am slowly mapping my FPGA audio rate cached casual dilated convolutions network to operate using their mat mul free GLU version of self attention. slowly getting there!
about 1 year ago
1
3
1
about 1 year ago
1
2
0
ooffff! the "DeepSeek-V3 Technical Report" is info dense!
arxiv.org/abs/2412.19437
loading . . .
DeepSeek-V3 Technical Report
We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepS...
https://arxiv.org/abs/2412.19437
about 1 year ago
1
2
0
still my fav "at doom's gate" cover! ( i'm probably 1K of those 2M views )
www.youtube.com/watch?v=TZG1...
loading . . .
DOOM: E1M1 (At Doom's Gate) - Metal Cover || RichaadEB & ToxicxEternity
YouTube video by RichaadEB
https://www.youtube.com/watch?v=TZG1eqPEYbY
about 1 year ago
0
0
0
this looks like it's going to be a good read!
www.amazon.com/AI-Robotics-...
loading . . .
AI for Robotics: Toward Embodied and General Intelligence in the Physical World
AI for Robotics: Toward Embodied and General Intelligence in the Physical World [Imran, Alishba, Gopalakrishnan, Keerthana] on Amazon.com. *FREE* shipping on qualifying offers. AI for Robotics: Toward Embodied and General Intelligence in the Physical World
https://www.amazon.com/AI-Robotics-Embodied-Intelligence-Physical/dp/B0DFLTZ4SQ
about 1 year ago
0
2
0
went camping in an area with lots of eucalyptus and was surrounded by koalas! so cute!
about 1 year ago
0
3
0
first logic gate waveshaper results tonight! this is network represented purely by combinational logic ( see
arxiv.org/abs/2210.08277
) i.e. just look up tables; no mults or adds. trained in jax ( 3 nested levels of vmap! ) & inference in amaranth. inference speed should be ~10s nanoseconds.
about 1 year ago
2
7
0
my largest category of splurge spend for last month was "bandcamp" :D
#winning
about 1 year ago
0
3
0
and here i am trying to debug why param_set((2,3,16), seed=1) vs param_set((2,3,16), seed=2) isn't working... ::/
about 1 year ago
1
1
0
by a lot of measures this was a completely irrational purchase, but, you know what i like it! it's ok to be irrational some times.
about 1 year ago
0
2
0
started learning "Amaranth HDL" this morning. my prior FPGA neural net stuff was verilog but switching to amaranth since the match the gateware on a device i'm targetting. as someone with a python background it's 1e6 times nicer than verilog!!
about 1 year ago
1
4
0
this board looks really fun! "a small, low cost board with the Raspberry Pi Pico processor (RP2040) and a Lattice Semiconductor iCE40UP5K FPGA"
pico-ice.tinyvision.ai
about 1 year ago
0
12
3
"Newton Losses: Using Curvature Information for Learning with Differentiable Algorithms". interesting approach using 2nd order info to avoid the exploding/vanishing gradients that come from non differentiable relaxations..
www.youtube.com/watch?v=7aFP...
loading . . .
Newton Losses: Using Curvature Information for Learning with Differentiable Algorithms - NeurIPS2024
YouTube video by Felix Petersen
https://www.youtube.com/watch?v=7aFPIMqSEpw
about 1 year ago
0
16
6
this is super cool!! i'll definitely be porting my fpga wavenet (
matpalm.com/blog/wavenet...
) in the new year
add a skeleton here at some point
about 1 year ago
1
6
0
still the best tech speech ever. "whatever path you're on, i advocate for long, hard work." so, so true. knowledge & skills do not just pop into our heads, we have to push and shove and jam it in ourselves.
www.youtube.com/watch?v=YOZn...
loading . . .
John Carmack Receives Honorary Degree
YouTube video by UMKC School of Science and Engineering
https://youtu.be/YOZnqjHkULc?si=YX85NjSLYl80QxcF&t=134
about 1 year ago
1
6
0
i always have this weird compulsion to move back to seattle. no idea why, given how bad the weather was lol
about 1 year ago
1
2
0
reposted by
Mat Kelcey
Edge Impulse
about 1 year ago
Today STMicroelectronics @ST_World announced the incredible new STM32N6, a powerhouse microcontroller with built-in neural processing that achieves 600x performance over current high-end STM32 MCUs — and offers native Edge Impulse support. Press release:
www.businesswire.com/news/home/20...
loading . . .
Edge Impulse and STMicroelectronics Launching Official Support of Revolutionary STM32N6 for Powerful Edge AI Solutions
Edge Impulse, the leading platform for building, deploying, and scaling edge machine learning models, and STMicroelectronics (NYSE: STM), a global sem
https://www.businesswire.com/news/home/20241210041857/en/Edge-Impulse-and-STMicroelectronics-Launching-Official-Support-of-Revolutionary-STM32N6-for-Powerful-Edge-AI-Solutions
1
2
2
love the arpeggiator on the oxi one, so playable and fun!
#improv
#eurorack
#techno
www.youtube.com/watch?v=mbpF...
loading . . .
2024 12 08 08 58 26 improv eurorack techno
YouTube video by dithernet
https://www.youtube.com/watch?v=mbpFr6_epIU
about 1 year ago
1
7
0
my 2025 goal is to play a hardware only improv techno set in a full club have played a few times in smaller settings around friends but hearing one of my tracks recently on a proper club PA has really inspired me!
dithernet.bandcamp.com/album/chapte...
loading . . .
chapter twenty two, by Dithernet
12 track album
https://dithernet.bandcamp.com/album/chapter-twenty-two
about 1 year ago
0
3
0
i love seeing people ( myself included ) rediscover old things. the recent flood of interest in AI which, for a lot of people, is 100% LLMs, is the best case i've seen of it! it really helps me solidify my only internal concepts to see someone else explain a decades old idea through new terminology.
about 1 year ago
1
3
0
"the new iris dataset" i'd almost forgotten about this!
about 1 year ago
0
2
0
the dog always waits. the gate in spring is shaded. she will run like hell.
#haiku
about 1 year ago
0
4
0
i can't repost this hard enough...
add a skeleton here at some point
over 1 year ago
0
2
0
keras3 vmapped jax models to tflite via orbax
matpalm.com/blog/keras3_...
over 1 year ago
0
3
0
sadly none of it "just works" with a keras3 jax model ( exported as stateless_function )
#sigh
no doubt i've just "simply" got my `jax_module.methods[constants.DEFAULT_METHOD_KEY]` wrong or some other magic arg combo, but not sure i can be bothered today...
add a skeleton here at some point
over 1 year ago
2
1
0
stuck a jax laptop sticker up in a nightclub bathroom friday night. hoping some spaced out dev has a 3am WTF moment...
over 1 year ago
1
8
1
#showYourSetup
#eurorack
over 1 year ago
2
27
2
(this week) jax to tflite is done with orbax ( nightly build only)
ai.google.dev/edge/litert/...
( and that's orbax-export, not orbax-checkpoint, since it's been split ) not sure what'll be next week, we'll see!
#canYouKeepUp
loading . . .
Jax Model Conversion For TFLite | Google AI Edge | Google AI for Developers
https://ai.google.dev/edge/litert/models/jax_to_tflite
over 1 year ago
0
3
1
whenever i see someone talk about a small 1,000,000,000 param LLM i think about the 20,000 param conv nets we've got deployed on microcontrollers. TBH am glad to not be in the LLM space right now
over 1 year ago
1
14
0
reposted by
Mat Kelcey
Daniel Situnayake
over 1 year ago
Edge AI is maturing, and we’re getting pretty good at sensory perception. What comes next? In this post I explore what we’ll need to bring real intelligence to the field. I’d love to hear what you think 👇🏽
open.substack.com/pub/dansitu/...
loading . . .
Integration
Edge AI has come a long way, but sensory perception is not enough. What will it take to build systems that can act?
https://open.substack.com/pub/dansitu/p/integration?utm_source=app-post-stats-page&r=b68xx&utm_medium=ios
0
7
6
have a number of old blog posts that are still useful but have code that no longer works... i know i should update them, just hard to be motivated to do so :/
#lazy
over 1 year ago
1
0
0
oh! just had an idea! in my fpga waveshaper (
matpalm.com/blog/wavenet...
) the limiting factor for speed was the ADAC which ran at 196kHz ( 5 micro second inference ) ...🧵
loading . . .
https://matpalm.com/blog/wavenet_on_fpga/the
over 1 year ago
1
2
0
updated my yolz blog post ( you only look zero shot times )
matpalm.com/blog/yolz/
still scrappy and hard to read i think but my time boxed hacking allocation has run up. onto something else!
loading . . .
brain of mat kelcey...
https://matpalm.com/blog/yolz/
over 1 year ago
0
2
0
it's time to celebrate when you reach 6D inputs tensors !
over 1 year ago
1
3
0
sometimes a jax.vmap can cause a MASSIVE blow out in jit time... i wonder what tools are available to understand why? ( it's not really profiling and jaxprs get pretty complex pretty quick... )
over 1 year ago
1
0
0
reposted by
Mat Kelcey
Shawn Hymel
over 1 year ago
The latest edition of Make Magazine is out and includes an article by yours truly 🤓 I talk about the collaboration project with @Odd_Jayy and give you steps to create your own local
#LLM
chatbot on a
#RaspberryPi
. Check it out:
www.makershed.com/collections/...
#maker
#ai
#electronics
loading . . .
3
38
3
it's _so_ important to write throwaway sampling debug tools. here's one for my zero shot object detector with top row being best performers and bottom row being worse. you never quite know what little quirk you'll pick up reviewing these...
loading . . .
over 1 year ago
0
3
1
after a couple of months of crappy weather it's about time to get jumping again!
loading . . .
over 1 year ago
1
2
0
latest small drone setup. super simple but fun.
over 1 year ago
0
3
0
the AQT lib
github.com/google/aqt/
is one of the few libs that does full quantised training ( not just quantised _aware_ training ) but it's implemented via a substitution of the dot_product? isn't jax all about fn transforms? why can't this "just" be a rewrite with tracing? i.e. `quantise(fn)`
over 1 year ago
0
1
0
Load more
feeds!
log in