Update README.md

README.md

@@ -4,7 +4,7 @@ language:
 license: gpl
 ---
 
-# Diversiformer
+# Diversiformer 🤗 🏳️‍🌈 🇩🇪
 
 _Work in progress._
 
@@ -31,31 +31,11 @@ Source code for fine-tuning is available [on GitHub](https://github.com/diversif
 ## Usage
 
 ```python
-from transformers import T5Tokenizer, TFT5ForConditionalGeneration
-
-tokenizer = T5Tokenizer.from_pretrained("diversifix/diversiformer")
-model = TFT5ForConditionalGeneration.from_pretrained("diversifix/diversiformer")
-
-def generate(prompt, tokenizer, model):
-    tokenized_text = tokenizer.encode(prompt, return_tensors="tf")
-    ids = model.generate(tokenized_text, max_length=500)
-    output = tokenizer.decode(ids[0], skip_special_tokens=True)
-    return output
-
-prompts = [
-    'Ersetze "Schüler" durch "Schülerin oder Schüler": Die Schüler kamen zu spät.',
-    'Ersetze "Lehrer" durch "Kollegium": Die wartenden Lehrer wunderten sich.',
-]
-
-for prompt in prompts:
-    output = generate(prompt, tokenizer, model)
-    print(f"{prompt}\n{output}\n\n")
+>>> from transformers import pipeline
+>>> generator = pipeline("text2text-generation", model="diversifix/diversiformer")
+>>> generator('Ersetze "Schüler" durch "Schülerin oder Schüler": Die Schüler kamen zu spät.', max_length=500)
 ```
 
-## Dev ideas
-
-- Use classifier to filter out training data of low quality? (~adversarial approach)
-
 ## License
 
 Diversiformer. Transformer model for inclusive language.