texcat-generation_review (#13)

- bug: correct generation parameters (2a1af1a85c406f402dbfec69ed1bdd13e0aa5066)

src/distilabel_dataset_generator/apps/base.py

@@ -427,6 +427,7 @@ def push_dataset_to_hub(
 
     if task == TEXTCAT_TASK:
         if num_labels == 1:
+            dataframe["label"] = dataframe["label"].replace("", None)
             features = Features(
                 {"text": Value("string"), "label": ClassLabel(names=labels)}
             )

src/distilabel_dataset_generator/apps/textcat.py

@@ -53,6 +53,9 @@ def push_dataset_to_hub(
     num_labels: int = 1,
 ):
     original_dataframe = dataframe.copy(deep=True)
+    dataframe = dataframe[
+        (dataframe["text"].str.strip() != "") & (dataframe["text"].notna())
+    ]
     labels = get_preprocess_labels(labels)
     try:
         push_to_hub_base(
@@ -80,6 +83,9 @@ def push_dataset_to_argilla(
     labels: List[str] = None,
 ) -> pd.DataFrame:
     original_dataframe = dataframe.copy(deep=True)
+    dataframe = dataframe[
+        (dataframe["text"].str.strip() != "") & (dataframe["text"].notna())
+    ]
     try:
         progress(0.1, desc="Setting up user and workspace")
         client = get_argilla_client()

src/distilabel_dataset_generator/pipelines/textcat.py

@@ -114,9 +114,11 @@ with Pipeline(name="textcat") as pipeline:
                 "temperature": 0.8,
                 "max_new_tokens": 2048,
                 "do_sample": True,
-                "seed": random.randint(0, 2**32 - 1),
+                "top_k": 50,
+                "top_p": 0.95,
             }},
         ),
+        seed=random.randint(0, 2**32 - 1),
         difficulty={None if difficulty == "mixed" else repr(difficulty)},
         clarity={None if clarity == "mixed" else repr(clarity)},
         num_generations={num_rows},
@@ -182,11 +184,13 @@ def get_textcat_generator(difficulty, clarity, is_sample):
                 "temperature": 0.9,
                 "max_new_tokens": 256 if is_sample else 2048,
                 "do_sample": True,
-                "seed": random.randint(0, 2**32 - 1),
+                "top_k": 50,
+                "top_p": 0.95,
             },
         ),
         difficulty=None if difficulty == "mixed" else difficulty,
         clarity=None if clarity == "mixed" else clarity,
+        seed=random.randint(0, 2**32 - 1),
     )
     textcat_generator.load()
     return textcat_generator

src/distilabel_dataset_generator/utils.py

@@ -124,4 +124,4 @@ def get_argilla_client() -> Union[rg.Argilla, None]:
         return None
 
 def get_preprocess_labels(labels: Optional[List[str]]) -> List[str]:
-    return [label.lower().strip() for label in labels] if labels else []
+    return list(set([label.lower().strip() for label in labels])) if labels else []