BlueHeeler-12M / README.md
mike-ravkine's picture
Update README.md
a8ce399
|
raw
history blame
211 Bytes
---
license: mit
language:
- en
---
BlueHeeler-10M is a nanoGPT (GPT-2) 6-head x 6-layer x 192-deep model trained on scripts from the children's show Bluey.
`iter 2000: loss 1.2913, time 30647.72ms, mfu 0.05%`