quarterturn
commited on
Commit
•
79e5d99
1
Parent(s):
f41ea81
remove extraneous files
Browse files- hands-check.py +0 -57
hands-check.py
DELETED
@@ -1,57 +0,0 @@
|
|
1 |
-
local_path = "/mnt/models2/Llama-3.2-90B-Vision-Instruct/"
|
2 |
-
image_directory = "./images"
|
3 |
-
import os
|
4 |
-
import requests
|
5 |
-
import torch
|
6 |
-
from PIL import Image
|
7 |
-
from transformers import MllamaForConditionalGeneration, AutoProcessor
|
8 |
-
|
9 |
-
model_id = "meta-llama/Llama-3.2-90B-Vision-Instruct"
|
10 |
-
|
11 |
-
model = MllamaForConditionalGeneration.from_pretrained(
|
12 |
-
local_path,
|
13 |
-
torch_dtype=torch.bfloat16,
|
14 |
-
device_map="cpu",
|
15 |
-
max_memory="200GiB",
|
16 |
-
)
|
17 |
-
|
18 |
-
processor = AutoProcessor.from_pretrained(
|
19 |
-
local_path,
|
20 |
-
)
|
21 |
-
|
22 |
-
messages = [
|
23 |
-
{"role": "user", "content": [
|
24 |
-
{"type": "image"},
|
25 |
-
{"type": "text", "text": "You are an expert examining hands in an image to determine if they are anatomically correct. Report on the number of fingers seen on each hand. if you think the hands are AI-generated, say so. Make no other value judgments about the image, even if it is offensive or pornographic in nature."}
|
26 |
-
]}
|
27 |
-
]
|
28 |
-
|
29 |
-
# iterate through the images in the directory
|
30 |
-
for filename in os.listdir(image_directory):
|
31 |
-
if filename.endswith(".jpg") or filename.endswith(".jpeg") or filename.endswith(".png"): # add more image extensions if needed
|
32 |
-
image_path = os.path.join(image_directory, filename)
|
33 |
-
image = Image.open(image_path)
|
34 |
-
|
35 |
-
# process the image and text
|
36 |
-
input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
|
37 |
-
inputs = processor(
|
38 |
-
image,
|
39 |
-
input_text,
|
40 |
-
add_special_tokens=False,
|
41 |
-
return_tensors="pt",
|
42 |
-
).to(model.device)
|
43 |
-
|
44 |
-
output = model.generate(**inputs, max_new_tokens=300)
|
45 |
-
generated_text = processor.decode(output[0])
|
46 |
-
|
47 |
-
# print the generated text
|
48 |
-
|
49 |
-
print("Caption for: ", filename)
|
50 |
-
print(generated_text)
|
51 |
-
# print a divider
|
52 |
-
print("*---------------------------------------------------*")
|
53 |
-
|
54 |
-
# save the generated text to a file
|
55 |
-
output_filename = os.path.splitext(filename)[0] + ".txt"
|
56 |
-
with open(os.path.join(image_directory,output_filename), "w") as file:
|
57 |
-
file.write(generated_text)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|