GPT2 model trained from scratch on the winemag reviews dataset to generate wine descriptions using text-generation. Note that these descriptions are mostly random descriptions.

Model description

How to use

You can use this model directly like so..

python

>>> from transformers import pipeline
>>> clm = pipeline('text-generation', model='panigrah/winberto-gpt2')
>>> clm("California Cabernet is", max_length=30, num_return_sequences=3)

[{'generated_text': 'California Pinot is a dark golden color. black plum and cherry aromas and flavors show their aromatic flair amidst ripe black fruit, cola and'},
 {'generated_text': 'California Pinot is a wine made from a grape that was aged in large oak tanks. the fruit is balanced by acidity and a crisp'},
 {'generated_text': 'California Pinot is a great surprise at all levels of age, but this delivers a soft, supple and luscious feel on the palate.'}]```

Training data

The GPT2 model was trained from scratch on 150K wine review descriptions. The training was cut short due at 5 epochs due to resource issues and still has a relatively high training and validatioan loss. The model is able to generate passable wine descriptions but they are not well correlated to the type of wine provided at the prompt itself.