curious how the token/$ on trainuim compare to nvidia chips and amd. is the future "chipless" inference?
Tokens per minute is a strange metric, I usually see tok/s. A per-second figure would be more accessible in comparing this to alternatives.
how does this compare with sglang
500 Code(3)
Error: Code(3)
curious how the token/$ on trainuim compare to nvidia chips and amd. is the future "chipless" inference?
Tokens per minute is a strange metric, I usually see tok/s. A per-second figure would be more accessible in comparing this to alternatives.
how does this compare with sglang