small code updates

app copy.py

@@ -9,10 +9,6 @@ import spaces
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 
-from chat_interface_preference import ChatInterface
-
-MAX_MAX_NEW_TOKENS = 2048
-DEFAULT_MAX_NEW_TOKENS = 1024
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "8192"))
 
 if torch.cuda.is_available():
@@ -63,7 +59,7 @@ def generate(
         yield "".join(outputs)
 
 
-chat_interface = ChatInterface(
+chat_interface = gr.ChatInterface(
     fn=generate,
     chatbot=gr.Chatbot(
         height=450, label="GEITje-SPIN", show_share_button=True, avatar_images=(None, "geitje-logo.jpg")

app.py

@@ -29,6 +29,7 @@ if torch.cuda.is_available():
     model_id = "Qwen/Qwen2-0.5B-Instruct"
     model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")
     tokenizer = AutoTokenizer.from_pretrained(model_id)
+
 style = "<style>.user-message,.system-message{display:flex;margin:10px}.user-message .message-content{background-color:#c2e3f7;color:#000}.system-message .message-content{background-color:#f5f5f5;color:#000}.message-content{padding:10px;border-radius:10px;max-width:70%;word-wrap:break-word}.container{display:flex;justify-content:space-between}.column{width:48%}</style>"
 
 client = rg.Argilla(api_url="https://davidberenstein1957-argilla-gradio.hf.space", api_key="owner.apikey")

chat_interface_preference.py

@@ -23,8 +23,9 @@ from gradio.components import (
     get_component_instance,
 )
 from gradio.events import Dependency, on
-from gradio.helpers import Error, Info, Warning, special_args
+from gradio.helpers import Error, Info, Warning
 from gradio.helpers import create_examples as Examples  # noqa: N812
+from gradio.helpers import special_args
 from gradio.layouts import Accordion, Group, Row
 from gradio.routes import Request
 from gradio.themes import ThemeClass as Theme
@@ -629,10 +630,6 @@ class ChatInterface(Blocks):
         n_generations: int = 1,
         *args,
     ) -> AsyncGenerator:
-        _, response = history_with_input[-1]
-        if self._check_if_two_responses(response):
-            raise Error("Two options detected: undo, log or random pick continuation.")
-
         if self.multimodal and isinstance(message, dict):
             remove_input = len(message["files"]) + 1 if message["text"] is not None else len(message["files"])
             history = history_with_input[:-remove_input]
@@ -646,14 +643,13 @@ class ChatInterface(Blocks):
                 generator = self.fn(*inputs)
             else:
                 generator = await anyio.to_thread.run_sync(self.fn, *inputs, limiter=self.limiter)
-                generator = SyncToAsyncIterator(generator, self.limiter)
+            generator = SyncToAsyncIterator(generator, self.limiter)
             first_response = await async_iteration(generator)
             if n_generations == 2:
                 first_response_formatted = self._get_chat_message_comparison(first_response, "")
             if self.multimodal and isinstance(message, dict):
                 for x in message["files"]:
                     history.append([(x,), None])
-
                 update = history + [[message["text"], first_response_formatted]]
                 yield update, update
             else:
@@ -670,10 +666,10 @@ class ChatInterface(Blocks):
             if n_generations == 2:
                 response_formatted = self._get_chat_message_comparison(response, "")
             if self.multimodal and isinstance(message, dict):
-                update = history + [[message["text"], response_formatted]]
+                update = history + [[message["text"], response]]
                 yield update, update
             else:
-                update = history + [[message, response_formatted]]
+                update = history + [[message, response]]
                 yield update, update
 
         if n_generations == 2:

test.py

@@ -1,8 +1,6 @@
 import random
 
 
-import argilla as rg
-
 from chat_interface_preference import ChatInterface