“Fine-tuning LoRa, rank 16, on 2,000 rows per arm, one epoch… 50 answers to each of 16 held-out questions at temperature 1, giving 800 answers per run.”
♫
Said on LessWrong (Curated & Popular)
Pitch The PM · EP. 55: Token Price Indexes Are Falling on Model Shifts Not Price Cuts with Carmen Li, CEO of Silicon Data & Compute Exchange
Audio streams from the publisher’s own feed. Verbatim does not host or copy it.
license data to cloud providers via exclusive or pay-and-share agreements, charging per data row (250,000 rows daily) or per transaction volume
250,000 data rows per day
“Fine-tuning LoRa, rank 16, on 2,000 rows per arm, one epoch… 50 answers to each of 16 held-out questions at temperature 1, giving 800 answers per run.”
“For things like the D statistic and the MAD score, you're really typically looking for about 80 samples, 80 observations, 80 rows of data.”
“You own the transaction data… you can license from them as well… we give you things that… our implied forward curve, our residual value.”
“We’re talking about the sock gen CTA index… we pick the 20 largest… give you daily data… close to 27 years of daily data”
The strongest new moments on your topics, in the speakers' own words. Free for three topics.
No card. No password. One-click unsubscribe.
Pro watches up to 25 names or phrases and emails you the morning after they are said.
See Pro, $29 a monthVerbatim quotes a short extract and links to the episode; the episode belongs to its publisher. Remove a show.