metadata

license: mit
datasets:
  - DarwinAnim8or/DMV-Plate-Review
language:
  - en
pipeline_tag: text-generation
tags:
  - dmv
  - fun
widget:
  - text: 'PLATE: LCDR'
    example_title: Plate LCDR
  - text: 'PLATE: LUCH'
    example_title: Plate LUCH
  - text: 'PLATE: JJ BINKS'
    example_title: Plate JJ BINKS
co2_eq_emissions:
  emissions: 20
  source: https://mlco2.github.io/impact/#compute
  training_type: fine-tuning
  geographical_location: Oregon, USA
  hardware_used: 1 T4, Google Colab

GPT-DMV-125m

A finetuned version of GPT-Neo-125M on the 'DMV' dataset. (Linked above) A demo is available here

(I recommend using the demo playground rather than the Inference window on the right here)

Training Procedure

This was trained on the 'DMV' dataset, using the "HappyTransformers" library on Google Colab. This model was trained for 5 epochs with learning rate 1e-2.

Biases & Limitations

This likely contains the same biases and limitations as the original GPT-Neo-125M that it is based on, and additionally heavy biases from the DMV dataset.

Intended Use

This model is meant for fun, nothing else.

Sample Use

#Import model:
from happytransformer import HappyGeneration
happy_gen = HappyGeneration("GPT-NEO", "DarwinAnim8or/GPT-DMV-125m")

#Set generation settings:
from happytransformer import GENSettings
args_top_k = GENSettings(no_repeat_ngram_size=3, do_sample=True,top_k=80, temperature=0.4, max_length=50, early_stopping=False)

#Generate a response:
result = happy_gen.generate_text("""PLATE: LUCH
REVIEW REASON CODE: """, args=args_top_k)

print(result)
print(result.text)