Spaces:
Running
on
Zero
Running
on
Zero
File size: 2,854 Bytes
30ef1d4 1557b00 3c739a1 1557b00 30ef1d4 01090a1 30ef1d4 01090a1 30ef1d4 3c73224 7a9fdcc 3c739a1 7a9fdcc 30ef1d4 7a9fdcc 30ef1d4 3c73224 30ef1d4 33cf987 01090a1 33cf987 d56d00d |
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 |
import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import os
import spaces
import torch
device = "cuda:0" if torch.cuda.is_available() else "cpu"
print(f'[INFO] Using device: {device}')
# token
token = os.environ['TOKEN']
# Load the pretrained model and tokenizer
MODEL_NAME = "BounharAbdelaziz/Al-Atlas-LLM-0.5B" # "atlasia/Al-Atlas-LLM-mid-training" # "BounharAbdelaziz/Al-Atlas-LLM-0.5B" #"atlasia/Al-Atlas-LLM"
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME) # , token=token
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME).to(device)
# Fix tokenizer padding
if tokenizer.pad_token is None:
tokenizer.pad_token = tokenizer.eos_token # Set pad token
# Predefined examples
examples = [
["الذكاء الاصطناعي هو فرع من علوم الكمبيوتر اللي كيركز"
, 256, 0.7, 0.9, 150, 8, 1.5],
["المستقبل ديال الذكاء الصناعي فالمغرب"
, 256, 0.7, 0.9, 150, 8, 1.5],
[" المطبخ المغربي"
, 256, 0.7, 0.9, 150, 8, 1.5],
["الماكلة المغربية كتعتبر من أحسن الماكلات فالعالم"
, 256, 0.7, 0.9, 150, 8, 1.5],
]
@spaces.GPU
def generate_text(prompt, max_length=256, temperature=0.7, top_p=0.9, top_k=150, num_beams=8, repetition_penalty=1.5):
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
output = model.generate(
**inputs,
max_length=max_length,
temperature=temperature,
top_p=top_p,
do_sample=True,
repetition_penalty=repetition_penalty,
num_beams=num_beams,
top_k= top_k,
early_stopping = True,
pad_token_id=tokenizer.pad_token_id, # Explicit pad token
eos_token_id=tokenizer.eos_token_id, # Explicit eos token
)
return tokenizer.decode(output[0], skip_special_tokens=True)
if __name__ == "__main__":
# Create the Gradio interface
with gr.Blocks() as app:
gr.Interface(
fn=generate_text,
inputs=[
gr.Textbox(label="Prompt: دخل النص بالدارجة"),
gr.Slider(8, 4096, value=256, label="Max Length"),
gr.Slider(0.0, 2, value=0.7, label="Temperature"),
gr.Slider(0.0, 1.0, value=0.9, label="Top-p"),
gr.Slider(1, 10000, value=150, label="Top-k"),
gr.Slider(1, 20, value=8, label="Number of Beams"),
gr.Slider(0.0, 100.0, value=1.5, label="Repetition Penalty"),
],
outputs=gr.Textbox(label="Generated Text in Moroccan Darija"),
title="Moroccan Darija LLM",
description="Enter a prompt and get AI-generated text using our pretrained LLM on Moroccan Darija.",
examples=examples,
)
app.launch()
|