contemmcm/qwen2.5-vl-3b-cls-full-bluesky-moderation

import json

from urllib.request import urlopen

import torch

from huggingface_hub import hf_hub_download
from PIL import Image
from transformers import AutoProcessor
from transformers.modeling_layers import GenericForSequenceClassification
from transformers.models.qwen2_5_vl.modeling_qwen2_5_vl import Qwen2_5_VLPreTrainedModel

MODEL = "contemmcm/qwen2.5-vl-3b-cls-full-bluesky-moderation"

# must be exactly this text: the model was trained with it
SYSTEM_PROMPT = (
    "You are a content moderation classifier for social media posts. "
    "Read the post and assess which moderation labels apply."
)


class Qwen2_5_VLForSequenceClassification(
    GenericForSequenceClassification, Qwen2_5_VLPreTrainedModel
):
    """transformers ships no classification head for Qwen2.5-VL. This is its generic one
    (a linear layer over the last token) on the Qwen2.5-VL backbone, which is the class
    the checkpoint was trained as."""


# The weights are stored under Qwen2.5-VL's original names. Without this mapping the
# language model is silently left randomly initialized and the scores are noise.
KEY_MAPPING = {
    "^visual": "model.visual",
    r"^model(?!\.(language_model|visual))": "model.language_model",
}

processor = AutoProcessor.from_pretrained(MODEL)
model = Qwen2_5_VLForSequenceClassification.from_pretrained(
    MODEL, dtype=torch.bfloat16, device_map="cuda:0", key_mapping=KEY_MAPPING
).eval()
labels = json.load(open(hf_hub_download(MODEL, "label_space.json")))["labels"]
thresholds = json.load(open(hf_hub_download(MODEL, "thresholds.json")))["thresholds"]

# A real post: https://bsky.app/profile/did:plc:3htuuatm2fchjvon7tpc2jge/post/3mtd2ffcsvk22
text = (
    "Get Your Summer Tees Here & Please FOLLOW & SHARE. Tees Starting At Just $16. (Wait for "
    "the 35-40% off sales, 2-3 times a month.) I, Also, Have Mugs, Tote Bags, Magnets, "
    "Stickers, Etc., Available. Over 250 Designs! If You DO Make A Purchase, Thanks In "
    "Advance! www.teepublic.com/user/roszelle-art"
)
image = Image.open(urlopen(
    "https://cdn.bsky.app/img/feed_fullsize/plain/did:plc:3htuuatm2fchjvon7tpc2jge/"
    "bafkreigp5m25icqkxlrm6pvbpzvey2g5bchrqonlyaypo7cs6ahqsekaey"
)).convert("RGB")

# Qwen spends visual tokens in proportion to resolution, and the model was trained on
# images whose longest side is at most 896 pixels. Larger ones must be shrunk first.
image.thumbnail((896, 896), Image.BILINEAR)

messages = [
    {"role": "system", "content": [{"type": "text", "text": SYSTEM_PROMPT}]},
    {"role": "user", "content": [{"type": "image", "image": image},
                                 {"type": "text", "text": text}]},
]
inputs = processor.apply_chat_template(
    messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt"
).to(model.device)

with torch.no_grad():
    probs = torch.sigmoid(model(**inputs).logits[0].float()).tolist()

# a label applies when its probability reaches that label's threshold; none means "safe"
for name, p, cut in zip(labels, probs, thresholds):
    if p >= cut:
        print(f"{name}: {p:.3f}")

spam: 0.534