File size: 4,210 Bytes
f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 b1e40ab f17f776 0aa5cd7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 |
import inspect
from typing import get_type_hints, Callable, Any
import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM
# --- Load Model and Tokenizer ---
model_id = "unsloth/SmolLM2-135M-Instruct-GGUF"
filename = "SmolLM2-135M-Instruct-Q8_0.gguf"
tokenizer = AutoTokenizer.from_pretrained(model_id, gguf_file=filename)
model = AutoModelForCausalLM.from_pretrained(model_id, gguf_file=filename)
# --- System Prompt Template ---
SYSTEM_PROMPT = """You are a helpful AI assistant. Your job is to provide clear and concise responses based on the user's input.
Keep your answers straightforward and avoid unnecessary information."""
def parse_docstring(func):
doc = inspect.getdoc(func)
if not doc:
return {"title": "Untitled", "description": ""}
lines = doc.splitlines()
title = next((line.replace("Title:", "").strip() for line in lines if line.startswith("Title:")), "Untitled")
description = "\n".join(line.strip() for line in lines if line.startswith("Description:"))
description = description.replace("Description:", "").strip()
return {"title": title, "description": description}
def gradio_app_with_docs(func: Callable) -> Callable:
sig = inspect.signature(func)
type_hints = get_type_hints(func)
metadata = parse_docstring(func)
"""
A decorator that automatically builds and launches a Gradio interface
based on function type hints.
Args:
func: A callable with type-hinted parameters and return type.
Returns:
The wrapped function with a `.launch()` method to start the app.
"""
def _map_type(t: type) -> gr.Component:
if t == str:
return gr.Textbox(label="Input")
elif t == int:
return gr.Number(precision=0)
elif t == float:
return gr.Number()
elif t == bool:
return gr.Checkbox()
elif hasattr(t, "__origin__") and t.__origin__ == list:
elem_type = t.__args__[0]
if elem_type == str:
return gr.Dropdown(choices=["Option1", "Option2"])
else:
raise ValueError(f"Unsupported list element type: {elem_type}")
else:
raise ValueError(f"Unsupported type: {t}")
# Build inputs
inputs = []
for name, param in sig.parameters.items():
if name == "self":
continue
param_type = type_hints.get(name, Any)
component = _map_type(param_type)
component.label = name.replace("_", " ").title()
inputs.append(component)
# Build outputs
return_type = type_hints.get("return", Any)
outputs = _map_type(return_type)
# Wrap function with Gradio interface
with gr.Blocks() as demo:
gr.Markdown(f"## {metadata['title']}\n{metadata['description']}")
gr.Interface(fn=func, inputs=inputs, outputs=outputs)
def wrapper(*args, **kwargs):
return func(*args, **kwargs)
wrapper.launch = lambda: demo.launch()
return wrapper
@gradio_app_with_docs
def generate_response(prompt: str) -> str:
"""
Title: Super Tiny GGUF Model on CPU
Description: A Simple app to test out the potentials of small GGUF LLM model.
Args:
prompt (str): A simple prompt.
Returns:
str: Simplified response.
"""
# Apply system prompt + user input
# full_prompt = f"<|begin_of_text|>System: {SYSTEM_PROMPT}\nUser: {prompt}\nAssistant:"
# inputs = tokenizer(full_prompt, return_tensors="pt").to("cpu")
messages = [
{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
enable_thinking=True # Switches between thinking and non-thinking modes. Default is True.
)
inputs = tokenizer([text], return_tensors="pt").to(model.device)
outputs = model.generate(
**inputs,
max_new_tokens=100,
# temperature=0.7,
# top_p=0.9
)
return tokenizer.decode(outputs[0], skip_special_tokens=True)
if __name__ == "__main__":
generate_response.launch() |