File size: 975 Bytes
9a70cb3 923ab5c 9a70cb3 923ab5c 9a70cb3 77bf2f9 183d759 9a70cb3 618804a 9a70cb3 a325776 9a70cb3 f790223 9a70cb3 61f602a 9a70cb3 ca1c3d5 2838c66 f790223 |
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 |
import gradio as gr
from fastai.vision.all import *
from huggingface_hub import from_pretrained_fastai
import torch, os
os.environ.setdefault("OMP_NUM_THREADS", "1")
torch.set_num_threads(1)
learn = from_pretrained_fastai("Pablogps/castle-classifier-25")
try:
learn.to_fp32()
except:
pass
labels = learn.dls.vocab
def predict(img):
img = PILImage.create(img) # same flow as before
pred, pred_idx, probs = learn.predict(img)
return {labels[i]: float(probs[i]) for i in range(len(labels))}
title = "Bad castle predictor"
description = "A bad model that tries to identify the type of castle."
examples = ['spanish', 'french', 'japanese']
demo = gr.Interface(
fn=predict,
inputs=gr.Image(type="pil"),
outputs=gr.Label(num_top_classes=3),
title=title,
description=description,
examples=examples,
cache_examples=False, # <-- don’t pre-run at startup
)
demo.queue(max_size=8).launch(show_error=True, debug=True) |