Update app.py

app.py

@@ -5,34 +5,30 @@ from collections.abc import Iterator
 from threading import Thread
 
 import gradio as gr
-import spaces
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 
 #
-# 1) Custom Pastel Gradient CSS
+# 1) Custom Pastel Gradient CSS, and force text to black
 #
 CUSTOM_CSS = """
 .gradio-container {
     background: linear-gradient(to right, #FFDEE9, #B5FFFC);
+    color: black; /* ensure text appears in black */
 }
 """
 
 #
-# 2) Description: Add French greeting, plus any info
+# 2) Description: "Bonjour Dans le chat du consentement" in black
+#    Also add a CPU notice in black if no GPU is found.
 #
-DESCRIPTION = """# Bonjour Dans le chat du consentement
-
-Mistral-7B Instruct Demo
+DESCRIPTION = """# Bonjour Dans le chat du consentement  
+Mistral-7B Instruct Demo  
 """
 
 if not torch.cuda.is_available():
-    DESCRIPTION += (
-        "\n<p style='color:red;'>Running on CPU - This is likely too large to run effectively.</p>"
-    )
+    DESCRIPTION += "Running on CPU - This is likely too large to run effectively.\n"
 
-MAX_MAX_NEW_TOKENS = 2048
-DEFAULT_MAX_NEW_TOKENS = 1024
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 #
@@ -41,8 +37,8 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 if torch.cuda.is_available():
     model_id = "mistralai/Mistral-7B-Instruct-v0.3"
     tokenizer = AutoTokenizer.from_pretrained(
-        model_id, 
-        trust_remote_code=True  # Might be needed for custom code
+        model_id,
+        trust_remote_code=True
     )
     model = AutoModelForCausalLM.from_pretrained(
         model_id,
@@ -54,106 +50,61 @@ if torch.cuda.is_available():
 def generate(
     message: str,
     chat_history: list[dict],
-    max_new_tokens: int = 1024,
-    temperature: float = 0.6,
-    top_p: float = 0.9,
-    top_k: int = 50,
-    repetition_penalty: float = 1.2,
 ) -> Iterator[str]:
     """
-    This function handles streaming chat text as the model generates it.
-    Uses Mistral's 'apply_chat_template' to handle chat-style prompting.
+    Minimal chat generation function: no sliders, no extra params.
     """
     conversation = [*chat_history, {"role": "user", "content": message}]
 
+    # Convert conversation to tokens
     input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
+    # If it exceeds max token length, trim
     if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
         input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
-        gr.Warning(
-            f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens."
-        )
+        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
+
     input_ids = input_ids.to(model.device)
 
+    # Use a streamer to yield tokens as they are generated
     streamer = TextIteratorStreamer(
-        tokenizer, 
-        timeout=20.0, 
-        skip_prompt=True, 
+        tokenizer,
+        timeout=20.0,
+        skip_prompt=True,
         skip_special_tokens=True
     )
+
+    # Basic generation settings (feel free to adjust if you want)
     generate_kwargs = dict(
-        {"input_ids": input_ids},
+        input_ids=input_ids,
         streamer=streamer,
-        max_new_tokens=max_new_tokens,
+        max_new_tokens=512,        # Adjust if you want more or fewer tokens
         do_sample=True,
-        top_p=top_p,
-        top_k=top_k,
-        temperature=temperature,
-        num_beams=1,
-        repetition_penalty=repetition_penalty,
+        temperature=0.7,
+        top_p=0.9,
+        repetition_penalty=1.1,
     )
+
+    # Run generation in a background thread
     t = Thread(target=model.generate, kwargs=generate_kwargs)
     t.start()
 
     outputs = []
     for text in streamer:
         outputs.append(text)
-        # Stream partial output as it's generated
         yield "".join(outputs)
 
 #
-# 4) Build the Chat Interface with extra sliders
+# 4) Build the Chat Interface
+#    - No additional sliders
+#    - No pre-filled example questions
 #
 demo = gr.ChatInterface(
     fn=generate,
     description=DESCRIPTION,
-    css=CUSTOM_CSS,  # Use our pastel gradient
-    additional_inputs=[
-        gr.Slider(
-            label="Max new tokens",
-            minimum=1,
-            maximum=MAX_MAX_NEW_TOKENS,
-            step=1,
-            value=DEFAULT_MAX_NEW_TOKENS,
-        ),
-        gr.Slider(
-            label="Temperature",
-            minimum=0.1,
-            maximum=4.0,
-            step=0.1,
-            value=0.6,
-        ),
-        gr.Slider(
-            label="Top-p (nucleus sampling)",
-            minimum=0.05,
-            maximum=1.0,
-            step=0.05,
-            value=0.9,
-        ),
-        gr.Slider(
-            label="Top-k",
-            minimum=1,
-            maximum=1000,
-            step=1,
-            value=50,
-        ),
-        gr.Slider(
-            label="Repetition penalty",
-            minimum=1.0,
-            maximum=2.0,
-            step=0.05,
-            value=1.2,
-        ),
-    ],
-    stop_btn=None,
-    examples=[
-        ["Hello there! How are you doing?"],
-        ["Can you explain briefly what the Python programming language is?"],
-        ["Explain the plot of Cinderella in a sentence."],
-        ["How many hours does it take a man to eat a Helicopter?"],
-        ["Write a 100-word article on 'Benefits of Open-Source in AI research'"],
-    ],
-    type="messages",
+    css=CUSTOM_CSS,
+    examples=None,  # remove example prompts
+    type="messages"
 )
 
 if __name__ == "__main__":
-    demo.queue(max_size=20).launch()
+    demo.queue().launch()