2 ms·
He mentions specifically that for machine learning its performance was underwhelming. Now he mentions tensor flow which probably has very different requirement
by wkat4242 4y ago
He mentions specifically that for machine learning its performance was underwhelming.
Now he mentions tensor flow which probably has very different requirements than running an LLM. But the point about the toolchain is a good one IMO.
Also, running an LLM is pretty niche for now since there are no open models yet (meta's is only an unofficial leak)