# Load model directly
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("textattack/bert-base-cased-STS-B")
model = AutoModelForSequenceClassification.from_pretrained("textattack/bert-base-cased-STS-B")Quick Links
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
TextAttack Model Card
This `bert-base-cased` model was fine-tuned for sequence classificationusing TextAttack
and the glue dataset loaded using the `nlp` library. The model was fine-tuned
for 3 epochs with a batch size of 128, a learning
rate of 1e-05, and a maximum sequence length of 128.
Since this was a regression task, the model was trained with a mean squared error loss function.
The best score the model achieved on this task was 0.8244429996636282, as measured by the
eval set pearson correlation, found after 2 epochs.
For more information, check out [TextAttack on Github](https://github.com/QData/TextAttack).
- Downloads last month
- 7
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="textattack/bert-base-cased-STS-B")