2 ms·
It's not really 100k years. A ROCm subset good enough to run an LLM can be done within a year or so. You can _technicall_ run it now with tinygrad. The silence
by cyberax 2y ago
It's not really 100k years. A ROCm subset good enough to run an LLM can be done within a year or so. You can _technicall_ run it now with tinygrad.
The silence from AMD has been deafening, though. I can't fathom why they're just ignoring the AI market.