BAAI
/

Edit model card

Generative Multimodal Models are In-Context Learners

Generative Multimodal Models are In-Context Learners

Quan Sun1*, Yufeng Cui1*, Xiaosong Zhang1*, Fan Zhang1*, Qiying Yu2,1*, Zhengxiong Luo1, Yueze Wang1, Yongming Rao1,
Jingjing Liu2, Tiejun Huang1,3, Xinlong Wang1†

1 BAAI, 2 THU, 3 PKU
* equal contribution † project lead

| Paper | πŸ€—HF Demo | Demo | Project Page | Github

The human ability to easily solve multimodal tasks in context (i.e., with only a few demonstrations or simple instructions), is what current multimodal systems have largely struggled to imitate. In this work, we demonstrate that the task-agnostic in-context learning capabilities of large multimodal models can be significantly enhanced by effective scaling-up. We introduce Emu2, a generative multimodal model with 37 billion parameters, trained on large-scale multimodal sequences with a unified autoregressive objective. Emu2 exhibits strong multimodal in-context learning abilities, even emerging to solve tasks that require on-the-fly reasoning, such as visual prompting and object-grounded generation. The model sets a new record on multiple multimodal understanding tasks in few-shot settings. When instruction-tuned to follow specific instructions, Emu2 further achieves new state-of-the-art on challenging tasks such as question answering benchmarks for large multimodal models and open-ended subject-driven generation. These achievements demonstrate that Emu2 can serve as a base model and general-purpose interface for a wide range of multimodal tasks. Code and models are publicly available to facilitate future research.

Model Weights

Model name Weight
Emu2 πŸ€— HF link
Emu2-Chat πŸ€— HF link
Emu2-Gen πŸ€— HF link

Inference (Huggingface Version)

Single GPU

from PIL import Image
import requests
import torch 
from transformers import AutoModelForCausalLM, AutoTokenizer


tokenizer = AutoTokenizer.from_pretrained("BAAI/Emu2")

model = AutoModelForCausalLM.from_pretrained(
    "BAAI/Emu2",
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
    trust_remote_code=True).to('cuda').eval()


# `[<IMG_PLH>]` is the image placeholder which will be replaced by image embeddings. 
# the number of `[<IMG_PLH>]` should be equal to the number of input images

query = '[<IMG_PLH>]Describe the image in details:' 
image = Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/blue_black_1_top_left.jpg?raw=true',stream=True).raw).convert('RGB')


inputs = model.build_input_ids(
    text=[query],
    tokenizer=tokenizer,
    image=[image]
)

with torch.no_grad():
     outputs = model.generate(
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],
        image=inputs["image"].to(torch.bfloat16),
        max_new_tokens=64,
        length_penalty=-1)

output_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)

Interleaved image and text

from PIL import Image
import requests
import torch 
from transformers import AutoModelForCausalLM, AutoTokenizer


tokenizer = AutoTokenizer.from_pretrained("BAAI/Emu2")

model = AutoModelForCausalLM.from_pretrained(
    "BAAI/Emu2",
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
    trust_remote_code=True).to('cuda').eval()

# `[<IMG_PLH>]` is the image placeholder which will be replaced by image embeddings. 
# the number of `[<IMG_PLH>]` should be equal to the number of input images

query = "[<IMG_PLH>][red, white, 3, bottom left].[<IMG_PLH>][yellow, white, 2, top left].[<IMG_PLH>][green, black, 4, bottom right][<IMG_PLH>]"

images = [
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/red_white_3_bottom_left.jpg?raw=true',stream=True).raw).convert('RGB'),
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/yellow_white_2_top_right.jpg?raw=true',stream=True).raw).convert('RGB'),
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/green_black_4_bottom_right.jpg?raw=true',stream=True).raw).convert('RGB'),
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/blue_black_1_top_left.jpg?raw=true',stream=True).raw).convert('RGB'),
]

inputs = model.build_input_ids(
    text=[query],
    tokenizer=tokenizer,
    image=images

)

with torch.no_grad():
     outputs = model.generate(
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],
        image=inputs["image"].to(torch.bfloat16),
        max_new_tokens=64,
        length_penalty=-1)

output_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)

Multi GPU

from PIL import Image 
import requests
import torch 
from transformers import AutoModelForCausalLM, AutoTokenizer
from accelerate import init_empty_weights, infer_auto_device_map, load_checkpoint_and_dispatch

tokenizer = AutoTokenizer.from_pretrained("BAAI/Emu2")

with init_empty_weights():
     model = AutoModelForCausalLM.from_pretrained(
        "BAAI/Emu2",
        torch_dtype=torch.bfloat16,
        low_cpu_mem_usage=True,
        trust_remote_code=True)  

device_map = infer_auto_device_map(model, max_memory={0:'38GiB',1:'38GiB',}, no_split_module_classes=['Block','LlamaDecoderLayer'])  
# input and output logits should be on same device
device_map["model.decoder.lm.lm_head"] = 0

model = load_checkpoint_and_dispatch(
    model, 
    'local/path/to/hf/version/Emu2/model',
    device_map=device_map).eval()

# `[<IMG_PLH>]` is the image placeholder which will be replaced by image embeddings. 
# the number of `[<IMG_PLH>]` should be equal to the number of input images

query = '[<IMG_PLH>]Describe the image in details:' 
image = Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/blue_black_1_top_left.jpg?raw=true',stream=True).raw).convert('RGB')

inputs = model.build_input_ids(
    text=[query],
    tokenizer=tokenizer,
    image=[image]

)

with torch.no_grad():
     outputs = model.generate(
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],
        image=inputs["image"].to(torch.bfloat16),
        max_new_tokens=64,
        length_penalty=-1)

output_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)

Interleaved image and text

from PIL import Image 
import requests
import torch 
from transformers import AutoModelForCausalLM, AutoTokenizer
from accelerate import init_empty_weights, infer_auto_device_map, load_checkpoint_and_dispatch

tokenizer = AutoTokenizer.from_pretrained("BAAI/Emu2")

with init_empty_weights():
     model = AutoModelForCausalLM.from_pretrained(
        "BAAI/Emu2",
        torch_dtype=torch.bfloat16,
        low_cpu_mem_usage=True,
        trust_remote_code=True)  

device_map = infer_auto_device_map(model, max_memory={0:'38GiB',1:'38GiB',}, no_split_module_classes=['Block','LlamaDecoderLayer'])  
# input and output logits should be on same device
device_map["model.decoder.lm.lm_head"] = 0

model = load_checkpoint_and_dispatch(
    model, 
    'local/path/to/hf/version/Emu2/model',
    device_map=device_map).eval()

# `[<IMG_PLH>]` is the image placeholder which will be replaced by image embeddings. 
# the number of `[<IMG_PLH>]` should be equal to the number of input images
query = "[<IMG_PLH>][red, white, 3, bottom left].[<IMG_PLH>][yellow, white, 2, top left].[<IMG_PLH>][green, black, 4, bottom right][<IMG_PLH>]"

images = [
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/red_white_3_bottom_left.jpg?raw=true',stream=True).raw).convert('RGB'),
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/yellow_white_2_top_right.jpg?raw=true',stream=True).raw).convert('RGB'),
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/green_black_4_bottom_right.jpg?raw=true',stream=True).raw).convert('RGB'),
    Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/blue_black_1_top_left.jpg?raw=true',stream=True).raw).convert('RGB'),
]

inputs = model.build_input_ids(
    text=[query],
    tokenizer=tokenizer,
    image=images

)

with torch.no_grad():
     outputs = model.generate(
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],
        image=inputs["image"].to(torch.bfloat16),
        max_new_tokens=64,
        length_penalty=-1)

output_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)

Quantization

Check quantization guidance at transformers

from PIL import Image 
import requests
import torch 
from transformers import AutoModelForCausalLM, AutoTokenizer


tokenizer = AutoTokenizer.from_pretrained("BAAI/Emu2")

model = AutoModelForCausalLM.from_pretrained(
    "BAAI/Emu2",
    load_in_4bit=True,
    trust_remote_code=True, 
    bnb_4bit_compute_dtype=torch.float16).eval()

query = '[<IMG_PLH>]Describe the image in details:' 
image = Image.open(requests.get('https://github.com/baaivision/Emu/Emu2/examples/blue_black_1_top_left.jpg?raw=true',stream=True).raw).convert('RGB')

inputs = model.build_input_ids(
    text=[query],
    tokenizer=tokenizer,
    image=[image]

)

with torch.no_grad():
     outputs = model.generate(
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],
        image=inputs["image"].to(torch.float16), # should be torch.float16
        max_new_tokens=64,
        length_penalty=-1)

output_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)

Citation

If you find Emu2 useful for your research and applications, please consider starring this repository and citing:

@article{Emu2,
    title={Generative Multimodal Models are In-Context Learners}, 
    author={Quan Sun and Yufeng Cui and Xiaosong Zhang and Fan Zhang and Qiying Yu and Zhengxiong Luo and Yueze Wang and Yongming Rao and Jingjing Liu and Tiejun Huang and Xinlong Wang},
    publisher={arXiv preprint arXiv:2312.13286},
    year={2023},
}
Downloads last month
158
Inference Examples
Inference API (serverless) does not yet support model repos that contain custom code.

Spaces using BAAI/Emu2 2