Update README.md

README.md

@@ -13,12 +13,26 @@ language:
 # Sentiment Classification in Polish
 
 ```
+import numpy as np
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
 
+id2label = {0: "negative", 1: "neutral", 2: "positive"}
 tokenizer = AutoTokenizer.from_pretrained("Voicelab/herbert-base-cased-sentiment")
 
 model = AutoModelForSequenceClassification.from_pretrained("Voicelab/herbert-base-cased-sentiment")
 
+encoding = tokenizer(
+          input,
+          add_special_tokens=True,
+          return_token_type_ids=True,
+          truncation=True,
+          padding='max_length',
+          return_attention_mask=True,
+          return_tensors='pt',
+        ).to("cuda:0")
+output = model(**encoding).logits.to("cpu").detach().numpy()
+prediction = id2label[np.argmax(output)]
+
 ```