fix: stop sequences for textgen and add examples to pipeline definiton

src/distilabel_dataset_generator/apps/sft.py

@@ -89,14 +89,18 @@ def generate_dataset(
             "You can only generate a dataset with 5000 or fewer rows. Setting to 5000."
         )
 
-    if num_rows < 50:
+    if num_rows < 10:
         duration = 60
-    elif num_rows < 250:
-        duration = 300
+    elif num_rows < 30:
+        duration = 120
+    elif num_rows < 100:
+        duration = 240
+    elif num_rows < 300:
+        duration = 600
     elif num_rows < 1000:
-        duration = 500
+        duration = 1200
     else:
-        duration = 1000
+        duration = 2400
 
     result_queue = multiprocessing.Queue()
     p = multiprocessing.Process(
@@ -127,7 +131,7 @@ def generate_dataset(
             repo_id=repo_id,
             private=private,
             include_script=False,
-            token=oauth_token.token,
+            token=oauth_token,
         )
 
     # If not pushing to hub generate the dataset directly

src/distilabel_dataset_generator/pipelines/sft.py

@@ -227,7 +227,6 @@ def get_prompt_generation_step():
                 "temperature": 0.8,
                 "max_new_tokens": 2048,
                 "do_sample": True,
-                "stop_sequences": _STOP_SEQUENCES,
             },
         ),
         use_system_prompt=True,
@@ -243,7 +242,7 @@ if __name__ == "__main__":
             [
                 {
                     "system_prompt": PROMPT_CREATION_PROMPT,
-                    "instruction": DEFAULT_DATASET_DESCRIPTION,
+                    "instruction": DEFAULT_DATASET_DESCRIPTIONS[0],
                 }
             ]
         )