Spaces:
Sleeping
Sleeping
File size: 1,775 Bytes
5385753 ab265b9 5385753 ab265b9 5385753 61d5788 733c10f 5385753 733c10f 5385753 | 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 | from __future__ import annotations
import sys
from pathlib import Path
import gradio as gr
import torch
sys.path.insert(0, str(Path(__file__).parent / "src"))
from tiny_transformer.train import load_checkpoint
CHECKPOINT = Path("demo/tiny-transformer-demo.pt")
DEVICE = "cpu"
model, tokenizer = load_checkpoint(str(CHECKPOINT), device=DEVICE)
def generate_text(
prompt: str,
max_new_tokens: int,
temperature: float,
top_k: int,
) -> str:
if not prompt:
prompt = "\n"
encoded = tokenizer.encode(prompt)
idx = torch.tensor([encoded], dtype=torch.long, device=DEVICE)
out = model.generate(
idx,
max_new_tokens=max_new_tokens,
temperature=temperature,
top_k=top_k,
)
return tokenizer.decode(out[0].tolist())
with gr.Blocks(title="Tiny Transformer") as demo:
gr.Markdown("# Tiny Transformer")
with gr.Row():
with gr.Column():
prompt = gr.Textbox(value="To be", label="Prompt", lines=5)
max_new_tokens = gr.Slider(8, 240, value=120, step=1, label="New tokens")
temperature = gr.Slider(0.2, 1.5, value=0.35, step=0.05, label="Temperature")
top_k = gr.Slider(1, 30, value=3, step=1, label="Top-k")
button = gr.Button("Generate", variant="primary")
output = gr.Textbox(label="Output", lines=16)
gr.Examples(
examples=[
["To be", 120, 0.35, 3],
["Attention", 120, 0.35, 3],
["The model", 120, 0.35, 3],
],
inputs=[prompt, max_new_tokens, temperature, top_k],
)
button.click(
generate_text,
inputs=[prompt, max_new_tokens, temperature, top_k],
outputs=output,
)
if __name__ == "__main__":
demo.launch()
|