pages / symptom / n_9fc79d80ca756b01

symptom

a buyer of one model is served another

of model delivery (a .genie program) · the program model_delivery · CC-BY-4.0

sealed cases that reached it: 0 — sealed cases that reached this word (the store counts, never reads)

shown by

• model_delivery: Deliver one inference request as the model the buyer named, with the proof its venue gives, or refuse it without a charge

fixed by

• a card serves one row of models.cue; its listing is gpu:<model id>:<venue>

caused by

• a card was listed under a price class (llama-3.1-8b) while its venue served other weights