Skip to content

Instantly share code, notes, and snippets.

@mstatt
Created August 31, 2024 06:21
Show Gist options
  • Select an option

  • Save mstatt/d7888198b20b762958299fd8377f8f3a to your computer and use it in GitHub Desktop.

Select an option

Save mstatt/d7888198b20b762958299fd8377f8f3a to your computer and use it in GitHub Desktop.
import os
import cv2
import torch
from PIL import Image
from transformers import AutoProcessor ,AutoImageProcessor
from transformers import AutoModelForCausalLM
## -------------------------------------------------------------------------------------------------------------------
def run_phi3_model(image_path,prompt):
model_id = "microsoft/Phi-3-vision-128k-instruct"
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cuda", trust_remote_code=True, torch_dtype="auto",attn_implementation="flash_attention_2")
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
messages = [
{"role": "user", "content": "<|image_1|>\n"+prompt},
]
# I will be using local images
image = Image.open(image_path)
prompt = processor.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = processor(prompt, [image], return_tensors="pt").to("cuda:0")
generation_args = {
"max_new_tokens": 1024,
"temperature": 0.0,
"do_sample": False,
}
generate_ids = model.generate(**inputs, eos_token_id=processor.tokenizer.eos_token_id, **generation_args)
# remove input tokens
generate_ids = generate_ids[:, inputs['input_ids'].shape[1]:]
response = processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
del model
del processor
return response
## -------------------------------------------------------------------------------------------------------------------
image_path = "1.png"
prompt = "Identify the 4 most dominant colors in this image and return the hex values of these 4 dominant colors. Respond ONLY with a list of these hex values."
results = run_phi3_model(image_path,prompt)
print(results)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment