Update example to English->German

README.md

@@ -119,19 +119,24 @@ pip3 install -e comet_early_exit
 This model is described in the appendix in the paper.
 It is able to score even *incomplete* translations (i.e. prefixes of translations):
 ```python
+import comet_early_exit
 model = comet_early_exit.load_from_checkpoint(comet_early_exit.download_model("zouharvi/COMET-partial"))
 data = [
     {
-        "src": "Can I receive my food in 10 to 15 minutes?",
-        "mt": "Mohl bych",
+        "src": "Will I receive my food in 10 to 15 minutes?",
+        "mt": "Ich bekomme",
     },
     {
-        "src": "Can I receive my food in 10 to 15 minutes?",
-        "mt": "Mohl bych dostat jídlo",
+        "src": "Will I receive my food in 10 to 15 minutes?",
+        "mt": "Bekomme ich",
     },
     {
-        "src": "Can I receive my food in 10 to 15 minutes?",
-        "mt": "Mohl bych dostat jídlo během 10 či 15 minut?",
+        "src": "Will I receive my food in 10 to 15 minutes?",
+        "mt": "Bekomme ich mein Essen in",
+    },
+    {
+        "src": "Will I receive my food in 10 to 15 minutes?",
+        "mt": "Bekomme ich mein Essen in 10 bis 15 Minuten?",
     }
 ]
 model_output = model.predict(data, batch_size=8, gpus=1)
@@ -139,7 +144,7 @@ print("scores", model_output["scores"])
 ```
 Outputs (formatted):
 ```
-scores 89.18  86.52  89.20
+scores 89.58 88.99 89.94 89.94
 ```
 
 This model is based on the work [Early-Exit and Instant Confidence Translation Quality Estimation](http://arxiv.org/abs/2502.14429) which can be cited as: