Mat Kelcey
@matpalm.com
data nerd wannabe
https://matpalm.com/
https://matpalm.com/
i wish this was always an option...
March 16, 2025 at 12:13 AM
i wish this was always an option...
"In what part of the world do you currently live?" lol
February 2, 2025 at 11:21 PM
"In what part of the world do you currently live?" lol
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!
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!
February 2, 2025 at 10:13 PM
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!
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!
went camping in an area with lots of eucalyptus and was surrounded by koalas! so cute!
January 21, 2025 at 10:24 AM
went camping in an area with lots of eucalyptus and was surrounded by koalas! so cute!
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.
January 14, 2025 at 10:55 AM
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.
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... ::/
January 5, 2025 at 11:10 PM
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... ::/
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.
January 5, 2025 at 9:39 PM
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.
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
"a small, low cost board with the Raspberry Pi Pico processor (RP2040) and a Lattice Semiconductor iCE40UP5K FPGA"
pico-ice.tinyvision.ai
December 30, 2024 at 8:46 AM
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
"a small, low cost board with the Raspberry Pi Pico processor (RP2040) and a Lattice Semiconductor iCE40UP5K FPGA"
pico-ice.tinyvision.ai
"the new iris dataset"
i'd almost forgotten about this!
i'd almost forgotten about this!
December 3, 2024 at 6:14 AM
"the new iris dataset"
i'd almost forgotten about this!
i'd almost forgotten about this!
November 28, 2024 at 12:22 AM
keras3 vmapped jax models to tflite via orbax
matpalm.com/blog/keras3_...
matpalm.com/blog/keras3_...
November 25, 2024 at 3:25 AM
keras3 vmapped jax models to tflite via orbax
matpalm.com/blog/keras3_...
matpalm.com/blog/keras3_...
November 23, 2024 at 12:11 AM
it's time to celebrate when you reach 6D inputs tensors !
November 17, 2024 at 6:51 AM
it's time to celebrate when you reach 6D inputs tensors !
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...
November 17, 2024 at 12:27 AM
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...
after a couple of months of crappy weather it's about time to get jumping again!
November 16, 2024 at 5:07 AM
after a couple of months of crappy weather it's about time to get jumping again!
latest small drone setup. super simple but fun.
November 15, 2024 at 10:40 PM
latest small drone setup. super simple but fun.
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)`
November 14, 2024 at 5:09 AM
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)`
extending my zero shot object detector...
( an example on held out data )
given input image ( top center )
and examples of reference objects ( bottom center )
predictions on left, ground truth on right
( an example on held out data )
given input image ( top center )
and examples of reference objects ( bottom center )
predictions on left, ground truth on right
November 12, 2024 at 8:20 AM
extending my zero shot object detector...
( an example on held out data )
given input image ( top center )
and examples of reference objects ( bottom center )
predictions on left, ground truth on right
( an example on held out data )
given input image ( top center )
and examples of reference objects ( bottom center )
predictions on left, ground truth on right
almost time to quit the day job!
November 6, 2024 at 12:06 AM
almost time to quit the day job!
yep, looks like that's exactly what it does!
November 3, 2024 at 8:54 PM
yep, looks like that's exactly what it does!
