Skip to content

Instantly share code, notes, and snippets.

@marktellez
Last active September 12, 2024 18:17
Show Gist options
  • Save marktellez/31a9dc0bcee66c68db1811e87ffcfebc to your computer and use it in GitHub Desktop.
Save marktellez/31a9dc0bcee66c68db1811e87ffcfebc to your computer and use it in GitHub Desktop.
initial audio dataset work - 3 hours - $0
transcription check time - 1.5 hours (16c * 1.5) - $0.24
model training time - 1 hour ($2 * 1) - $2
generation prep time - 0.5 hours - $1
generation time (1000 msgs) (8 secs * 1000 msgs / 4 workers (unattended) (non blocking) - 1 hour - $2
generation check time (1000 messages) (20 secs to check * 2 checkers * 1000 msgs / 2 teams of 2) - 3 hours (16c * 3) - $0.48
cost - $5
Does not include transfer or s3 file storage or databases for simplicity and clarity
initial audio dataset work - 3 hours
transcription check time - 1.5 hours
model training time - 1 hour
generation prep time - 1 hour
generation time (1000 msgs) (8 secs * 1000 msgs / 4 workers (unattended) (non blocking) - 1 hour
generation check time (1000 messages) (20 secs to check * 2 checkers * 1000 msgs / 2 teams of 2) - 3 hours
cost - our time for various worker pays (+ inexpensive electricity)
investment - $2k GPU + $1k in Hardware
@marktellez
Copy link
Author

these are PER VOICE

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment