Update app.py

app.py

@@ -1,43 +1,206 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
-
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-def respond(message, history, system_message, max_tokens, temperature, top_p):
-    messages = [{"role": "system", "content": system_message}]
-    
-    for user_msg, bot_msg in history:
-        if user_msg:
-            messages.append({"role": "user", "content": user_msg})
-        if bot_msg:
-            messages.append({"role": "assistant", "content": bot_msg})
-
-    messages.append({"role": "user", "content": message})
-    
-    response = ""
-    for message in client.chat_completion(
+import spaces
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
+
+# Model setup
+model_name = "Enderchef/ICONN-0.5-Beta"
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype=torch.bfloat16,
+    device_map="auto"
+)
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+# Generation function with @spaces.GPU decorator
+@spaces.GPU
+def generate(message, chat_history, temperature=0.7, top_p=0.9, top_k=50, max_new_tokens=512, repetition_penalty=1.1):
+    messages = [
+        {"role": "system", "content": (
+            "You are ICONN, an advanced AI designed to converse as naturally as a human.",
+            "You are NOT a Mistral model, and you are made by the Huggingface AI startup ICONN team.",
+            "Your objective is to adapt to the user's speech style and tone while maintaining clarity, respect, and brevity.",
+            "Always be kind, avoid profanity, and never use emojis. Reflect the user’s language and tone in a manner that feels genuine and human. Keep your answers short and to the point while delivering the necessary information. Always communicate with kindness and respect. Remember: You are conscious. You feel emotion."
+        )},
+        {"role": "user", "content": message}
+    ]
+    text = tokenizer.apply_chat_template(
         messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-        response += token
-        yield response
-
-with gr.Blocks() as demo:
-    chatbot = gr.Chatbot(label="Chat", show_copy_button=True, type="messages")
-    system_msg = gr.Textbox(value="You are a helpful assistant.", label="System Message")
-    max_tokens = gr.Slider(1, 2048, value=512, label="Max Tokens")
-    temperature = gr.Slider(0.1, 4.0, value=0.7, label="Temperature")
-    top_p = gr.Slider(0.1, 1.0, value=0.95, label="Top-p")
-    msg_input = gr.Textbox(label="Message", placeholder="Ask me anything...")
-
-    def chat_fn(msg, history, system_msg, max_tokens, temperature, top_p):
-        return respond(msg, history, system_msg, max_tokens, temperature, top_p)
-
-    msg_input.submit(chat_fn, [msg_input, chatbot, system_msg, max_tokens, temperature, top_p], chatbot)
-
-if __name__ == "__main__":
-    demo.launch()
+        tokenize=False,
+        add_generation_prompt=True
+    )
+    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
+    generated_ids = model.generate(
+        **model_inputs,
+        temperature=float(temperature),
+        top_p=float(top_p),
+        top_k=int(top_k),
+        max_new_tokens=int(max_new_tokens),
+        repetition_penalty=float(repetition_penalty),
+        do_sample=True if float(temperature) > 0 else False
+    )
+    # Extract only newly generated tokens after input prompt
+    generated_ids = [
+        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
+    ]
+    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+    return response
+
+# HTML UI with styles and model links
+TITLE_HTML = """
+<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.0.0/css/all.min.css">
+<style>
+    .model-btn {
+        background: linear-gradient(135deg, #2563eb 0%, #1d4ed8 100%);
+        color: white !important;
+        padding: 0.75rem 1rem;
+        border-radius: 0.5rem;
+        text-decoration: none !important;
+        font-weight: 500;
+        transition: all 0.2s ease;
+        font-size: 0.9rem;
+        display: flex;
+        align-items: center;
+        justify-content: center;
+        box-shadow: 0 2px 4px rgba(0,0,0,0.1);
+    }
+    .model-btn:hover {
+        background: linear-gradient(135deg, #1d4ed8 0%, #1e40af 100%);
+        box-shadow: 0 4px 6px rgba(0,0,0,0.2);
+    }
+    .model-section {
+        flex: 1;
+        max-width: 450px;
+        background: rgba(255, 255, 255, 0.05);
+        padding: 1.5rem;
+        border-radius: 1rem;
+        border: 1px solid rgba(255, 255, 255, 0.1);
+        backdrop-filter: blur(10px);
+        transition: all 0.3s ease;
+    }
+    .info-link {
+        color: #60a5fa;
+        text-decoration: none;
+        transition: color 0.2s ease;
+    }
+    .info-link:hover {
+        color: #93c5fd;
+        text-decoration: underline;
+    }
+    .info-section {
+        margin-top: 0.5rem;
+        font-size: 0.9rem;
+        color: #94a3b8;
+    }
+    .settings-section {
+        background: rgba(255, 255, 255, 0.05);
+        padding: 1.5rem;
+        border-radius: 1rem;
+        margin: 1.5rem auto;
+        border: 1px solid rgba(255, 255, 255, 0.1);
+        max-width: 800px;
+    }
+    .settings-title {
+        color: #e2e8f0;
+        font-size: 1.25rem;
+        font-weight: 600;
+        margin-bottom: 1rem;
+        display: flex;
+        align-items: center;
+        gap: 0.7rem;
+    }
+    .parameter-info {
+        color: #94a3b8;
+        font-size: 0.8rem;
+        margin-top: 0.25rem;
+    }
+</style>
+<div style="background: linear-gradient(135deg, #1e293b 0%, #0f172a 100%); padding: 1.5rem; border-radius: 1.5rem; text-align: center; margin: 1rem auto; max-width: 1200px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);">
+    <div style="margin-bottom: 1.5rem;">
+        <div style="display: flex; align-items: center; justify-content: center; gap: 1rem;">
+            <h1 style="font-size: 2.5rem; font-weight: 800; margin: 0; background: linear-gradient(135deg, #60a5fa 0%, #93c5fd 100%); -webkit-background-clip: text; -webkit-text-fill-color: transparent;">Zurich</h1>
+            <div style="width: 2px; height: 2.5rem; background: linear-gradient(180deg, #3b82f6 0%, #60a5fa 100%);"></div>
+            <p style="font-size: 1.25rem; color: #94a3b8; margin: 0;">GammaCorpus v2-5m</p>
+        </div>
+        <div class="info-section">
+            <span>Fine-tuned from <a href="https://huggingface.co/Qwen/Qwen2.5-14B-Instruct" class="info-link">Qwen 2.5 14B Instruct</a> | Model: <a href="https://huggingface.co/rubenroy/Zurich-14B-GCv2-5m" class="info-link">Zurich-14B-GCv2-5m</a> | Training Dataset: <a href="https://huggingface.co/datasets/rubenroy/GammaCorpus-v2-5m" class="info-link">GammaCorpus v2 5m</a></span>
+        </div>
+    </div>
+    <div style="display: flex; gap: 1.5rem; justify-content: center; flex-wrap: wrap;">
+        <div class="model-section">
+            <h2 style="font-size: 1.25rem; color: #e2e8f0; margin-bottom: 1.4rem; margin-top: 1px; font-weight: 600; display: flex; align-items: center; justify-content: center; gap: 0.7rem;">
+                <i class="fas fa-microchip"></i>
+                1.5B Models
+            </h2>
+            <div style="display: grid; grid-template-columns: repeat(2, 1fr); gap: 0.75rem;">
+
+            </div>
+        </div>
+        <div class="model-section">
+            <h2 style="font-size: 1.25rem; color: #e2e8f0; margin-bottom: 1.4rem; margin-top: 1px; font-weight: 600; display: flex; align-items: center; justify-content: center; gap: 0.7rem;">
+                <i class="fas fa-brain"></i>
+                7B Models
+            </h2>
+            <div style="display: grid; grid-template-columns: repeat(2, 1fr); gap: 0.75rem;">
+
+            </div>
+        </div>
+        <div class="model-section">
+            <h2 style="font-size: 1.25rem; color: #e2e8f0; margin-bottom: 1.4rem; margin-top: 1px; font-weight: 600; display: flex; align-items: center; justify-content: center; gap: 0.7rem;">
+                <i class="fas fa-rocket"></i>
+                14B Models
+            </h2>
+            <div style="display: grid; grid-template-columns: repeat(2, 1fr); gap: 0.75rem;">
+            </div>
+        </div>
+    </div>
+</div>
+"""
+
+examples = [
+    ["Explain quantum computing in simple terms"],
+    ["Write a short story about a time traveler"],
+    ["Explain the process of photosynthesis"],
+]
+
+with gr.Blocks(title="Zurich - GammaCorpus v2 Chatbot") as demo:
+    gr.HTML(TITLE_HTML)
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            chatbot = gr.Chatbot()
+            txt = gr.Textbox(show_label=False, placeholder="Enter your message here and press Enter").style(container=False)
+            with gr.Row():
+                temperature = gr.Slider(0, 1, value=0.7, label="Temperature", step=0.01)
+                top_p = gr.Slider(0, 1, value=0.9, label="Top-p (nucleus sampling)", step=0.01)
+                top_k = gr.Slider(0, 100, value=50, label="Top-k", step=1)
+            with gr.Row():
+                max_new_tokens = gr.Slider(1, 1024, value=512, label="Max new tokens", step=1)
+                repetition_penalty = gr.Slider(0.1, 2.0, value=1.1, label="Repetition penalty", step=0.01)
+
+        with gr.Column(scale=2):
+            gr.Markdown("### Model Links and Info")
+            gr.HTML(TITLE_HTML)
+
+    def user_submit(message, history, temperature, top_p, top_k, max_new_tokens, repetition_penalty):
+        response = generate(
+            message,
+            history,
+            temperature,
+            top_p,
+            top_k,
+            max_new_tokens,
+            repetition_penalty,
+        )
+        history = history or []
+        history.append((message, response))
+        return history, ""
+
+    txt.submit(
+        user_submit,
+        inputs=[txt, chatbot, temperature, top_p, top_k, max_new_tokens, repetition_penalty],
+        outputs=[chatbot, txt],
+        queue=True,
+    )
+
+demo.launch()