12MTRANSFORMER / README.md
Roman190928's picture
Update README.md
ded1250 verified
metadata
license: mit
datasets:
  - starhopp3r/TinyChat
language:
  - en
pipeline_tag: text-generation
library_name: transformers
tags:
  - meh
  - bad
  - horrible
  - dont-use
  - skilless
  - 12m
  - good

total steps = 9250

total time(seconds) = 4650.91

final param count = 12,717,312

final loss = 1.287906

avg loss = 2.080128

took too long to train on 1xT4