11–20 of 127 posts
Earlier quoted context omitted.
> every implementation idea rediscovered five times and wrapped in a new README That's open source since forever, unfortunately.
It's what happens when you don't do market research.
There are already a handful of repos doing essentially exactly this: `mlx-moe-offload`, `streamlx`, `mlx-moe`, `mlx-flash`, and `deepseek-v4-flash-mlx` - i.e. keep the resident parts of an MoE in unified memory and page/stream routed experts from SSD on Apple Silicon. At this point I'd much rather see people collaborate on one of these implementations, benchmark against them, or upstream the useful bits into MLX/MLX-…
> every implementation idea rediscovered five times and wrapped in a new README That's open source since forever, unfortunately.
I genuinely want to contribute. And hey! I was doing oss this since 2014 so waay before AI was cool.
Earlier quoted context omitted.
It's what happens when you don't do market research.
I'm sorry this makes it seem like I didn't do my research. I did a TON. To fix it I'll add a benchmark/comparison table. Also, I wouldn't call it market research since this is not commercial AT ALL.
There are already a handful of repos doing essentially exactly this: `mlx-moe-offload`, `streamlx`, `mlx-moe`, `mlx-flash`, and `deepseek-v4-flash-mlx` - i.e. keep the resident parts of an MoE in unified memory and page/stream routed experts from SSD on Apple Silicon. At this point I'd much rather see people collaborate on one of these implementations, benchmark against them, or upstream the useful bits into MLX/MLX-…
It seems we could use a new kind of memory that streams the weight data in, like GDDR in reverse.
AI;DR
Earlier quoted context omitted.
> every implementation idea rediscovered five times and wrapped in a new README That's open source since forever, unfortunately.
It's what happens when you don't do market research.