philtor πΊπ¦π»
-
This week experimented with adding a matmul & sorting accelerators (and opcodes for both) to PicoRV32 using the PCPI interface. Also added support for those ops to the Spike RISCV simulator. Yes, Claude helped a huge amount with these tasks which would've likely taken weeks otherwise. The design simulates (Verilator) & sythesizes (yosys) and I can generate a bitstream for both GateMateA1 and ECP5 (OrangeCrab board)
-
Portland high temp tomorrow forecast to be only 20F. That would be the coldest high temperature in over 30 years. #pdxst βοΈ
-
@Zach777 certainly Disney is not great, but DeSantis is potentially worse. Nice to see DeSantis get his comeuppance #LetThemFight
(Also what DeSantis was trying to do wouldn't have helped Disney workers, he has no problem with companies taking advantage of workers)
-
OpenAI down? No internet? "Open"AI not so open anymore? You can run your own local LLM chat model called Alpaca. I got it going on my home machine today, while not as good as chatGPT (It's a much smaller model) it's still pretty cool to chat with and it's amazing to think it's running on my old PC and not even using a GPU. (kind of a weird answer to the fibonacci question there)
https://github.com/antimatter15/alpaca.cpp #LLM #chatbot #DIYAI
-
I need to dig into the paper more deeply, but this seems like a pretty significant result for optimizing ML performance and power usage (at least for classifiers - but it seems like the same approach could be used elsewhere).
"Despite Tiny Classifiers being constrained to a few hundred logic gates, we observe no statistically significant difference in prediction performance in comparison to the best-performing ML baseline"
-
-
Just found this QuickLogic EOS S3 FPGA devboard on sale for $32.18 on SparkFun. Going to give it a try. QuckLogic FPGAs are supported by open source SymbiFlow tools.