Spaces:
Sleeping
Sleeping
File size: 1,682 Bytes
08a4054 18a3215 08a4054 c87f74b 18a3215 08a4054 3856fba 579abe4 18a3215 08a4054 579abe4 f7aecd4 579abe4 f7aecd4 08a4054 c2ed9b2 506d5e7 08a4054 579abe4 150e6b4 579abe4 08a4054 |
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 |
import os
import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer
tokenizer = AutoTokenizer.from_pretrained("afrizalha/Bakpia-V1-0.5B-Javanese")
model = AutoModelForCausalLM.from_pretrained("afrizalha/Bakpia-V1-0.5B-Javanese")
text_streamer = TextStreamer(tokenizer)
template = """<|im_start|>system
<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
"""
def generate(query, temp, top_p):
inputs = template.format(prompt=query)
inputs = tokenizer([inputs], return_tensors="pt").to(model.device)
outputs = model.generate(
streamer = text_streamer,
inputs=inputs.input_ids,
max_new_tokens=1024,
do_sample=True,
temperature=temp,
top_p=top_p)
outputs = tokenizer.decode(outputs[0][input_length:], skip_special_tokens=True)
return outputs
with gr.Blocks(theme=gr.themes.Soft()) as app:
input = gr.Textbox(label="Prompt", value="Pripun kulo saged nyinaoni Basa Jawa kanthi sae?")
output = gr.Textbox(label="Response", scale=2)
temp = gr.Slider(label="Temperature", minimum=0.1, maximum=1.0, step=0.1, value=0.5)
top_p = gr.Slider(label="Top P", minimum=0.1, maximum=1.0, step=0.1, value=0.5)
gr.Interface(
fn=generate,
inputs=[input,temp,top_p],
outputs=[output],
allow_flagging="never",
title="Bakpia V1 0.5B",
description = """Bakpia V1 0.5B is a fine-tuned version of Qwen 2 0.5B Instruct. It is fine-tuned using massive synthetic data for Krama Javanese, where the prompts are generated by GPT-4o and the responses are generated by Claude 3 Haiku."""
)
app.launch() |