GPT2-RPGPT-8.48M / README.md
xzuyn's picture
Update README.md
f392c8f
metadata
language:
  - en
pipeline_tag: text-generation
datasets:
  - practicaldreamer/RPGPT_PublicDomain-alpaca

Latest Version: 10,000 Steps

  • 2,560,000 tokens seen.

Model Info:

  • Trained from scratch.
  • 8.5M parameters.
  • 256 context length.
  • Test model. Likely needs at least 512 context to function "properly".
  • Trained with a dataset that overlaps by a quarter of the context length (Shifts by 64 tokens for each subset).

Format:

<|characters|>
Nancy (Oliver Twist): Female, early 20s, ESFP, Cockney accent. Loyal...
Mr. Edward Hyde (Dr. Jekyll and Mr. Hyde): Male, late 30s, ESTP...

<|scenario|>
In an alternate Victorian London where the city's poor and downtrodden...

<|response|>
Nancy: *gently brushes her fingers across the worn book spine, before suddenly stopping as she feels another hand...
Mr. Edward Hyde: *glances at Nancy with a sinister grin, slowly pulling his hand back* No need to apologize, miss...

Example Output:

Step 10,000.

<|characters|>
The Intended (The Iliad): Male, late 20s, ENFJ, British accent. Intelligent, empathetic, seeking the same challenge by Jane Austen's challenges.

<|scenario|>
In a strange sense of a mysterious time, The Intended finds himself in the same place. The Red and the

Config:

batch_size: 1
dropout: 0
learning_rate: 0.0001
max_length: 256
n_embed: 256
n_head: 8
n_layer: 8
vocab_size: 8192