wip

convlab/policy/ussT5/emowoz_evaluate.py

@@ -65,7 +65,6 @@ def generate_result(model_checkpoint, data, stop=-1):
     for input_text, target_text in tqdm(zip(data["input_text"], data["target_text"]), ascii=True):
         if stop > 0 and i > stop:
             break
-        if "satisfaction score" in input_text:
         i += 1
         inputs = tokenizer([input_text], return_tensors="pt", padding=True)
         output = model.generate(input_ids=inputs["input_ids"],
@@ -76,9 +75,10 @@ def generate_result(model_checkpoint, data, stop=-1):
         if len(output) > 1:
             print(output)
             output = "illegal"
-
+        if "satisfaction score" in input_text:
+            output = tri_convert(output)
         results.append({"input_text": input_text,
-                            "preds": tri_convert(output),
+                        "preds": output,
                         "label": target_text})
     json.dump(results, open(os.path.join(
         model_checkpoint, "emowoz_result.json"), 'w'))
@@ -89,6 +89,7 @@ def read_result(result):
     preds = []
     label = []
     for r in result:
+        if "satisfaction score" in r["input_text"]:
             preds.append(r["preds"])
             label.append(r["label"])
     return preds, label