zlacker

Android adds ggml lib to AICore

submitted by tosh+(OP) on 2023-12-09 16:23:17 | 74 points 9 comments
[view article] [source] [links] [go to bottom]
replies(3): >>api+qi >>lern_t+Ml >>westur+gp1
◧
1. api+qi[view] [source] 2023-12-09 18:27:33
>>tosh+(OP)
Anyone else want to call this library “Gargamel” after the character from The Smurfs? I can’t unsee it.
replies(4): >>SushiH+Ul >>TeMPOr+9v >>satvik+dI >>kaoD+gK
◧
2. lern_t+Ml[view] [source] 2023-12-09 18:47:10
>>tosh+(OP)
The whole thread for those without an account: https://nitter.net/tarantulae/status/1733263854312825220
◧◩
3. SushiH+Ul[view] [source] [discussion] 2023-12-09 18:47:56
>>api+qi
Thank you! I always do in my head and I thought I was the only one
◧◩
4. TeMPOr+9v[view] [source] [discussion] 2023-12-09 19:35:01
>>api+qi
Thanks! I hope this becomes the (un)official common moniker - it rolls off the tongue better than "gee gee em el".
replies(1): >>cozzyd+XD
◧◩◪
5. cozzyd+XD[view] [source] [discussion] 2023-12-09 20:23:48
>>TeMPOr+9v
I like to think of it as "Good Game, Machine Learning"
replies(1): >>double+OI
◧◩
6. satvik+dI[view] [source] [discussion] 2023-12-09 20:46:33
>>api+qi
I call it Gaugamela after Alexander's famous rout there.

https://en.wikipedia.org/wiki/Battle_of_Gaugamela

◧◩◪◨
7. double+OI[view] [source] [discussion] 2023-12-09 20:49:45
>>cozzyd+XD
Same.
◧◩
8. kaoD+gK[view] [source] [discussion] 2023-12-09 21:00:12
>>api+qi
After years of gaming I can't help but see ggml as an initialism.
◧
9. westur+gp1[view] [source] 2023-12-10 02:50:52
>>tosh+(OP)
From "PyTorch for WebGPU" (2023) >>36009478 :

> Fwiw it looks like the llama.cpp Tensor is from ggml, for which there are CUDA and OpenCL implementations (but not yet ROCm, or a WebGPU shim for use with emscripten transpilation to WASM): https://github.com/ggerganov/llama.cpp/blob/master/ggml.h

[go to top]