tests/test_idefics2.py

import requests
import torch
from PIL import Image
from io import BytesIO

from transformers import AutoProcessor, AutoModelForVision2Seq, AwqConfig


MODE = "quantized"
DEVICE = "cuda"
PROCESSOR = AutoProcessor.from_pretrained("HuggingFaceM4/idefics2-tfrm-compatible")
BAD_WORDS_IDS = PROCESSOR.tokenizer(
    ["<image>", "<fake_token_around_image>"], add_special_tokens=False
).input_ids
EOS_WORDS_IDS = PROCESSOR.tokenizer(
    "<end_of_utterance>", add_special_tokens=False
).input_ids + [PROCESSOR.tokenizer.eos_token_id]

# Load model
if MODE == "regular":
    model = AutoModelForVision2Seq.from_pretrained(
        "HuggingFaceM4/idefics2-tfrm-compatible",
        torch_dtype=torch.float16,
        trust_remote_code=True,
        _attn_implementation="flash_attention_2",
        revision="3dc93be345d64fb6b1c550a233fe87ddb36f183d",
    ).to(DEVICE)
elif MODE == "quantized":
    quant_path = "HuggingFaceM4/idefics2-tfrm-compatible-AWQ"
    model = AutoModelForVision2Seq.from_pretrained(
        quant_path, trust_remote_code=True
    ).to(DEVICE)
elif MODE == "fused_quantized":
    quant_path = "HuggingFaceM4/idefics2-tfrm-compatible-AWQ"
    quantization_config = AwqConfig(
        bits=4,
        fuse_max_seq_len=4096,
        modules_to_fuse={
            "attention": ["q_proj", "k_proj", "v_proj", "o_proj"],
            "mlp": ["gate_proj", "up_proj", "down_proj"],
            "layernorm": ["input_layernorm", "post_attention_layernorm", "norm"],
            "use_alibi": False,
            "num_attention_heads": 32,
            "num_key_value_heads": 8,
            "hidden_size": 4096,
        },
    )
    model = AutoModelForVision2Seq.from_pretrained(
        quant_path,
        quantization_config=quantization_config,
        trust_remote_code=True,
    ).to(DEVICE)
else:
    raise ValueError("Unknown mode")


def download_image(url):
    try:
        # Send a GET request to the URL to download the image
        response = requests.get(url)
        # Check if the request was successful (status code 200)
        if response.status_code == 200:
            # Open the image using PIL
            image = Image.open(BytesIO(response.content))
            # Return the PIL image object
            return image
        else:
            print(f"Failed to download image. Status code: {response.status_code}")
            return None
    except Exception as e:
        print(f"An error occurred: {e}")
        return None


# Create inputs
image1 = download_image(
    "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
)


def ask_vlm(image, instruction):
    prompts = [
        "User:",
        image,
        f"{instruction}.<end_of_utterance>\n",
        "Assistant:",
    ]
    inputs = PROCESSOR(prompts)
    inputs = {k: torch.tensor(v).to(DEVICE) for k, v in inputs.items()}
    generated_ids = model.generate(
        **inputs,
        bad_words_ids=BAD_WORDS_IDS,
        max_new_tokens=100,
    )
    generated_texts = PROCESSOR.batch_decode(generated_ids, skip_special_tokens=True)
    return generated_texts


import time

model.eval()
now = time.time()
print(ask_vlm(image1, "What is this?")[0].split("\nAssistant: ")[1])

print("resp:", time.time() - now)
import time

now = time.time()

print(ask_vlm(image1, "What is this?")[0].split("\nAssistant: ")[1])