"""
STEP 3 — Gradio app that uses YOUR OWN model
=============================================
Upload this as app.py to your Hugging Face Space.
It loads YOUR model (dineshkasi/my-ai-assistant) — not anyone else's!
Your friends visit:
https://huggingface.co/spaces/YOUR_USERNAME/YOUR_MODEL_NAME
"""
import torch
import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
# ══════════════════════════════════════════════════════
# ✏️ CHANGE THESE TO YOUR OWN DETAILS
# ══════════════════════════════════════════════════════
HF_USERNAME = "DineshKasi"
MODEL_NAME = "ai-assistant"
# ══════════════════════════════════════════════════════
REPO_ID = f"{HF_USERNAME}/{MODEL_NAME}"
print(f"Loading YOUR model: {REPO_ID} ...")
tokenizer = AutoTokenizer.from_pretrained(REPO_ID)
model = AutoModelForCausalLM.from_pretrained(REPO_ID)
model.eval()
print("Model loaded!")
def ask(question, history, temperature, max_tokens):
"""Generate a response from YOUR model."""
# Build context from history (last 3 turns)
context = ""
for user_msg, bot_msg in history[-3:]:
context += f"<|user|>{user_msg}<|endoftext|>"
context += f"<|assistant|>{bot_msg}<|endoftext|>"
prompt = context + f"<|user|>{question}<|endoftext|><|assistant|>"
inputs = tokenizer.encode(prompt, return_tensors="pt")
with torch.no_grad():
output = model.generate(
inputs,
max_new_tokens=int(max_tokens),
temperature=float(temperature),
do_sample=True,
top_p=0.92,
repetition_penalty=1.1,
pad_token_id=tokenizer.eos_token_id,
eos_token_id=tokenizer.eos_token_id,
)
full_text = tokenizer.decode(output[0], skip_special_tokens=True)
# Extract only the last assistant reply
if "<|assistant|>" in full_text:
reply = full_text.split("<|assistant|>")[-1].strip()
else:
reply = full_text[len(prompt):].strip()
return reply
# ── Gradio UI ─────────────────────────────────────────
with gr.Blocks(
theme=gr.themes.Soft(primary_hue="violet"),
title=f"{HF_USERNAME}'s AI Assistant",
css="""
#header { text-align:center; padding: 24px 0 8px; }
#header h1 { font-size:2rem; font-weight:700; color:#7c3aed; margin:0; }
#header p { color:#6b7280; margin:6px 0 0; }
.badge {
display:inline-block; background:#f3e8ff; color:#7c3aed;
border-radius:999px; padding:3px 14px; font-size:0.8rem;
margin:3px; font-weight:500;
}
#badges { text-align:center; margin:8px 0 18px; }
#model-credit {
text-align:center; margin-top:14px;
font-size:0.8rem; color:#9ca3af;
}
#model-credit a { color:#7c3aed; text-decoration:none; }
footer { display:none !important; }
""",
) as demo:
gr.HTML(f"""
💻 Coding
🔬 Science
📐 Math
✍️ Writing
📊 Business
🌍 General
""")
chatbot = gr.Chatbot(
label="Chat with my model",
bubble_full_width=False,
height=460,
)
with gr.Row():
txt = gr.Textbox(
placeholder="Ask me anything...",
show_label=False,
scale=8,
container=False,
)
btn = gr.Button("Send ➤", variant="primary", scale=1)
gr.Examples(
examples=[
"Write a Python function to find prime numbers",
"Explain how neural networks learn",
"What is the difference between RAM and ROM?",
"Help me write a professional email",
"What is quantum entanglement?",
"How does a binary search tree work?",
],
inputs=txt,
label="💡 Try these",
)
with gr.Accordion("⚙️ Settings", open=False):
temperature = gr.Slider(0.1, 1.2, value=0.7, step=0.1,
label="Temperature (creativity)")
max_tokens = gr.Slider(64, 512, value=200, step=32,
label="Max response length")
with gr.Accordion("🔌 Use this model in your project", open=False):
gr.Markdown(f"""
### Load my model directly in Python
```python
pip install transformers torch
```
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
tokenizer = AutoTokenizer.from_pretrained("{REPO_ID}")
model = AutoModelForCausalLM.from_pretrained("{REPO_ID}")
def ask(question):
prompt = f"<|user|>{{question}}<|endoftext|><|assistant|>"
inputs = tokenizer.encode(prompt, return_tensors="pt")
with torch.no_grad():
output = model.generate(inputs, max_new_tokens=200,
temperature=0.7, do_sample=True,
pad_token_id=tokenizer.eos_token_id)
reply = tokenizer.decode(output[0], skip_special_tokens=True)
return reply.split("<|assistant|>")[-1].strip()
print(ask("Explain machine learning in simple terms"))
```
### Via Hugging Face Inference API (no install)
```python
from huggingface_hub import InferenceClient
client = InferenceClient("{REPO_ID}")
result = client.text_generation(
"<|user|>What is a neural network?<|endoftext|><|assistant|>",
max_new_tokens=200,
)
print(result)
```
> Model page: [huggingface.co/{REPO_ID}](https://huggingface.co/{REPO_ID})
""")
gr.HTML(f"""
Model:
{REPO_ID}
· Built by {HF_USERNAME}
· Hosted on Hugging Face Spaces