File size: 1,884 Bytes
8c1a99e a4a68e2 8c1a99e a4a68e2 8c1a99e |
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 |
# Importing necessary libraries
import subprocess
import spaces
from transformers import AutoProcessor, AutoModelForCausalLM
# Install the required dependencies
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
# Load model and processor from Hugging Face
model_id = "microsoft/Florence-2-large-ft"
model = (
AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True).to("cuda").eval()
)
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
@spaces.GPU(duration=120)
def run_example(task_prompt, image, text_input=None):
"""
Runs an example using the given task prompt and image.
Args:
task_prompt (str): The task prompt for the example.
image (PIL.Image.Image): The image to be processed.
text_input (str, optional): Additional text input to be appended to the task prompt. Defaults to None.
Returns:
str: The parsed answer generated by the model.
"""
# If there is no text input, use the task prompt as the prompt
if text_input is None:
prompt = task_prompt
else:
prompt = task_prompt + text_input
# Process the image and text input
inputs = processor(text=prompt, images=image, return_tensors="pt").to("cuda")
# Generate the answer using the model
generated_ids = model.generate(
input_ids=inputs["input_ids"],
pixel_values=inputs["pixel_values"],
max_new_tokens=1024,
early_stopping=False,
do_sample=False,
num_beams=3,
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
parsed_answer = processor.post_process_generation(
generated_text, task=task_prompt, image_size=(image.width, image.height)
)
# Return the parsed answer
return parsed_answer
|