Spaces:
Running
on
Zero
Running
on
Zero
File size: 1,483 Bytes
c207c47 |
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 |
from PIL import Image
import spaces
import gradio as gr
from transformers import (
AutoProcessor,
AutoModelForCausalLM,
)
import torch
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
Florence_models = AutoModelForCausalLM.from_pretrained(
"microsoft/Florence-2-large", torch_dtype=torch_dtype, trust_remote_code=True
).to(device)
Florence_processors = AutoProcessor.from_pretrained(
"microsoft/Florence-2-large", trust_remote_code=True
)
@spaces.GPU
def feifeiflorence(
image,
progress=gr.Progress(track_tqdm=True),
):
image = Image.fromarray(image)
task_prompt = "<MORE_DETAILED_CAPTION>"
if image.mode != "RGB":
image = image.convert("RGB")
inputs = Florence_processors(
text=task_prompt, images=image, return_tensors="pt"
).to(device, torch_dtype)
generated_ids = Florence_models.generate(
input_ids=inputs["input_ids"],
pixel_values=inputs["pixel_values"],
max_new_tokens=1024,
num_beams=3,
do_sample=False,
)
generated_text = Florence_processors.batch_decode(
generated_ids, skip_special_tokens=False
)[0]
parsed_answer = Florence_processors.post_process_generation(
generated_text, task=task_prompt, image_size=(image.width, image.height)
)
return parsed_answer["<MORE_DETAILED_CAPTION>"]
|