Visual-language assistant with InternVL2 and OpenVINO#

This Jupyter notebook can be launched after a local installation only.

InternVL 2.0 is the latest addition to the InternVL series of multimodal large language models. InternVL 2.0 features a variety of instruction-tuned models, ranging from 1 billion to 108 billion parameters. Compared to the state-of-the-art open-source multimodal large language models, InternVL 2.0 surpasses most open-source models. It demonstrates competitive performance on par with proprietary commercial models across various capabilities, including document and chart comprehension, infographics QA, scene text understanding and OCR tasks, scientific and mathematical problem solving, as well as cultural understanding and integrated multimodal capabilities.

More details about model can be found in model card, blog and original repo.

In this tutorial we consider how to convert and optimize InternVL2 model for creating multimodal chatbot. Additionally, we demonstrate how to apply stateful transformation on LLM part and model optimization techniques like weights compression using NNCF

Table of contents:

Prerequisites
Select model
Convert and Optimize model
- Compress model weights to 4-bit
Select inference device
Prepare model inference pipeline
Run model inference
Interactive demo

Installation Instructions#

This is a self-contained example that relies solely on its own code.

We recommend running the notebook in a virtual environment. You only need a Jupyter server to start. For details, please refer to Installation Guide.

Prerequisites#

%pip install -q "transformers>4.36,<4.45" "torch>=2.1" "torchvision" "einops" "timm" "Pillow" "gradio>=4.36" --extra-index-url https://download.pytorch.org/whl/cpu
%pip install -q "openvino>=2024.3.0" "nncf>=2.12.0"

Note: you may need to restart the kernel to use updated packages.
Note: you may need to restart the kernel to use updated packages.

from pathlib import Path
import requests

if not Path("conversation.py").exists():
    r = requests.get("https://huggingface.co/OpenGVLab/InternVL2-1B/raw/main/conversation.py")
    open("conversation.py", "w", encoding="utf-8").write(r.text)

if not Path("internvl2_helper.py").exists():
    r = requests.get(url="https://raw.githubusercontent.com/openvinotoolkit/openvino_notebooks/latest/notebooks/internvl2/internvl2_helper.py")
    open("internvl2_helper.py", "w", encoding="utf-8").write(r.text)

if not Path("gradio_helper.py").exists():
    r = requests.get(url="https://raw.githubusercontent.com/openvinotoolkit/openvino_notebooks/latest/notebooks/internvl2/gradio_helper.py")
    open("gradio_helper.py", "w", encoding="utf-8").write(r.text)

if not Path("notebook_utils.py").exists():
    r = requests.get(url="https://raw.githubusercontent.com/openvinotoolkit/openvino_notebooks/latest/utils/notebook_utils.py")
    open("notebook_utils.py", "w", encoding="utf-8").write(r.text)

Select model#

There are multiple InternVL2 models available in models collection. You can select one of them for conversion and optimization in notebook using widget bellow:

from internvl2_helper import model_selector

model_id = model_selector()

model_id

INFO:nncf:NNCF initialized successfully. Supported frameworks detected: torch, tensorflow, onnx, openvino

Dropdown(description='Model:', options=('OpenGVLab/InternVL2-1B', 'OpenGVLab/InternVL2-2B', 'OpenGVLab/InternV…

print(f"Selected {model_id.value}")
pt_model_id = model_id.value
model_dir = Path(pt_model_id.split("/")[-1])

Selected OpenGVLab/InternVL2-1B

Convert and Optimize model#

InternVL2 is PyTorch model. OpenVINO supports PyTorch models via conversion to OpenVINO Intermediate Representation (IR). OpenVINO model conversion API should be used for these purposes. ov.convert_model function accepts original PyTorch model instance and example input for tracing and returns ov.Model representing this model in OpenVINO framework. Converted model can be used for saving on disk using ov.save_model function or directly loading on device using core.complie_model. internvl2_helper.py script contains helper function for model conversion, please check its content if you interested in conversion details.

Click here for more detailed explanation of conversion steps InternVL2 is autoregressive transformer generative model, it means that each next model step depends from model output from previous step. The generation approach is based on the assumption that the probability distribution of a word sequence can be decomposed into the product of conditional next word distributions. In other words, model predicts the next token in the loop guided by previously generated tokens until the stop-condition will be not reached (generated sequence of maximum length or end of string token obtained). The way the next token will be selected over predicted probabilities is driven by the selected decoding methodology. You can find more information about the most popular decoding methods in this blog. The entry point for the generation process for models from the Hugging Face Transformers library is the generate method. You can find more information about its parameters and configuration in the documentation. To preserve flexibility in the selection decoding methodology, we will convert only model inference for one step.

The inference flow has difference on first step and for the next. On the first step, model accept preprocessed input instruction and image, that transformed to the unified embedding space using input_embedding and image_encoder models, after that language model, LLM-based part of model, runs on input embeddings to predict probability of next generated tokens. On the next step, language_model accepts only next token id selected based on sampling strategy and processed by input_embedding model and cached attention key and values. Since the output side is auto-regressive, an output token hidden state remains the same once computed for every further generation step. Therefore, recomputing it every time you want to generate a new token seems wasteful. With the cache, the model saves the hidden state once it has been computed. The model only computes the one for the most recently generated output token at each time step, re-using the saved ones for hidden tokens. This reduces the generation complexity from \(O(n^3)\) to \(O(n^2)\) for a transformer model. More details about how it works can be found in this article. To sum up above, model consists of 4 parts:

Image encoder for encoding input images into embedding space.
Input Embedding for conversion input text tokens into embedding space
Language Model for generation answer based on input embeddings provided by Image Encoder and Input Embedding models.

Compress model weights to 4-bit#

For reducing memory consumption, weights compression optimization can be applied using NNCF.

Click here for more details about weight compression Weight compression aims to reduce the memory footprint of a model. It can also lead to significant performance improvement for large memory-bound models, such as Large Language Models (LLMs). LLMs and other models, which require extensive memory to store the weights during inference, can benefit from weight compression in the following ways:

enabling the inference of exceptionally large models that cannot be accommodated in the memory of the device;
improving the inference performance of the models by reducing the latency of the memory access when computing the operations with weights, for example, Linear layers.

Neural Network Compression Framework (NNCF) provides 4-bit / 8-bit mixed weight quantization as a compression method primarily designed to optimize LLMs. The main difference between weights compression and full model quantization (post-training quantization) is that activations remain floating-point in the case of weights compression which leads to a better accuracy. Weight compression for LLMs provides a solid inference performance improvement which is on par with the performance of the full model quantization. In addition, weight compression is data-free and does not require a calibration dataset, making it easy to use.

nncf.compress_weights function can be used for performing weights compression. The function accepts an OpenVINO model and other compression parameters. Compared to INT8 compression, INT4 compression improves performance even more, but introduces a minor drop in prediction quality.

More details about weights compression, can be found in OpenVINO documentation.

from internvl2_helper import convert_internvl2_model

# uncomment these lines to see model conversion code
# convert_internvl2_model??

import nncf

compression_configuration = {
    "mode": nncf.CompressWeightsMode.INT4_ASYM,
    "group_size": 128,
    "ratio": 1.0,
}

convert_internvl2_model(pt_model_id, model_dir, compression_configuration)

⌛ OpenGVLab/InternVL2-1B conversion started. Be patient, it may takes some time.
⌛ Load Original model

/opt/home/k8sworker/ci-ai/cibuilds/jobs/ov-notebook/jobs/OVNotebookOps/builds/810/archive/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
  warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning)

FlashAttention2 is not installed.
✅ Original model successfully loaded
⌛ Convert Input embedding model
WARNING:nncf:NNCF provides best results with torch==2.4.*, while current torch version is 2.2.2+cpu. If you encounter issues, consider switching to torch==2.4.*
✅ Input embedding model successfully converted
⌛ Convert Image embedding model

/opt/home/k8sworker/ci-ai/cibuilds/jobs/ov-notebook/jobs/OVNotebookOps/builds/810/archive/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/transformers/modeling_utils.py:4713: FutureWarning: _is_quantized_training_enabled is going to be deprecated in transformers 4.39.0. Please use model.hf_quantizer.is_trainable instead
  warnings.warn(
/opt/home/k8sworker/.cache/huggingface/modules/transformers_modules/OpenGVLab/InternVL2-1B/a84c71e158b16180df4fd1c5fe963fdf54b2cd43/modeling_internvl_chat.py:195: TracerWarning: Converting a tensor to a Python integer might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  h = w = int(vit_embeds.shape[1] ** 0.5)
/opt/home/k8sworker/.cache/huggingface/modules/transformers_modules/OpenGVLab/InternVL2-1B/a84c71e158b16180df4fd1c5fe963fdf54b2cd43/modeling_internvl_chat.py:169: TracerWarning: Converting a tensor to a Python integer might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  x = x.view(n, w, int(h * scale_factor), int(c / scale_factor))
/opt/home/k8sworker/.cache/huggingface/modules/transformers_modules/OpenGVLab/InternVL2-1B/a84c71e158b16180df4fd1c5fe963fdf54b2cd43/modeling_internvl_chat.py:173: TracerWarning: Converting a tensor to a Python integer might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  x = x.view(n, int(h * scale_factor), int(w * scale_factor),
/opt/home/k8sworker/.cache/huggingface/modules/transformers_modules/OpenGVLab/InternVL2-1B/a84c71e158b16180df4fd1c5fe963fdf54b2cd43/modeling_internvl_chat.py:174: TracerWarning: Converting a tensor to a Python integer might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  int(c / (scale_factor * scale_factor)))

⌛ Weights compression with int4_asym mode started

2024-11-05 01:40:40.561024: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.
2024-11-05 01:40:40.599895: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-11-05 01:40:41.225225: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

INFO:nncf:Statistics of the bitwidth distribution:
┍━━━━━━━━━━━━━━━━┯━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┯━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┑
│   Num bits (N) │ % all parameters (layers)   │ % ratio-defining parameters (layers)   │
┝━━━━━━━━━━━━━━━━┿━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┿━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┥
│              8 │ 0% (2 / 99)                 │ 0% (0 / 97)                            │
├────────────────┼─────────────────────────────┼────────────────────────────────────────┤
│              4 │ 100% (97 / 99)              │ 100% (97 / 97)                         │
┕━━━━━━━━━━━━━━━━┷━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┷━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┙

Output()

✅ Weights compression finished
✅ Image embedding model successfully converted
⌛ Convert Language model
WARNING:tensorflow:Please fix your imports. Module tensorflow.python.training.tracking.base has been moved to tensorflow.python.trackable.base. The old module will be deleted in version 2.11.

We detected that you are passing past_key_values as a tuple and this is deprecated and will be removed in v4.43. Please use an appropriate Cache class (https://huggingface.co/docs/transformers/v4.41.3/en/internal/generation_utils#transformers.Cache)
/opt/home/k8sworker/ci-ai/cibuilds/jobs/ov-notebook/jobs/OVNotebookOps/builds/810/archive/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/transformers/models/qwen2/modeling_qwen2.py:100: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if sequence_length != 1:
/opt/home/k8sworker/ci-ai/cibuilds/jobs/ov-notebook/jobs/OVNotebookOps/builds/810/archive/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/transformers/models/qwen2/modeling_qwen2.py:165: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if seq_len > self.max_seq_len_cached:
/opt/home/k8sworker/ci-ai/cibuilds/jobs/ov-notebook/jobs/OVNotebookOps/builds/810/archive/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/transformers/models/qwen2/modeling_qwen2.py:324: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
/opt/home/k8sworker/ci-ai/cibuilds/jobs/ov-notebook/jobs/OVNotebookOps/builds/810/archive/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/transformers/models/qwen2/modeling_qwen2.py:339: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):

✅ Language model successfully converted
⌛ Weights compression with int4_asym mode started
INFO:nncf:Statistics of the bitwidth distribution:
┍━━━━━━━━━━━━━━━━┯━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┯━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┑
│   Num bits (N) │ % all parameters (layers)   │ % ratio-defining parameters (layers)   │
┝━━━━━━━━━━━━━━━━┿━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┿━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┥
│              8 │ 28% (1 / 169)               │ 0% (0 / 168)                           │
├────────────────┼─────────────────────────────┼────────────────────────────────────────┤
│              4 │ 72% (168 / 169)             │ 100% (168 / 168)                       │
┕━━━━━━━━━━━━━━━━┷━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┷━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┙

Output()

✅ Weights compression finished
✅ OpenGVLab/InternVL2-1B model conversion finished. You can find results in InternVL2-1B

Select inference device#

from notebook_utils import device_widget

device = device_widget(default="AUTO", exclude=["NPU"])

device

Dropdown(description='Device:', index=1, options=('CPU', 'AUTO'), value='AUTO')

Prepare model inference pipeline#

As discussed, the model comprises Image Encoder and LLM (with separated text embedding part) that generates answer. In internvl2_helper.py we defined LLM inference class OvModelForCausalLMWithEmb that will represent generation cycle, It is based on HuggingFace Transformers GenerationMixin and looks similar to Optimum Intel OVModelForCausalLM that is used for LLM inference with only difference that it can accept input embedding. In own turn, general multimodal model class OVInternVLChatModel handles chatbot functionality including image processing and answer generation using LLM.

from internvl2_helper import OVInternVLChatModel
from transformers import AutoTokenizer

# Uncomment below lines to see the model inference class code

# OVInternVLChatModel??

tokenizer = AutoTokenizer.from_pretrained(model_dir, trust_remote_code=True)
ov_model = OVInternVLChatModel(model_dir, device.value)

Run model inference#

Our interface is fully compatible with Transformers interface for InternVL2, you can try any of represented here usage examples. Let’s check model capabilities in answering questions about image:

import PIL
from internvl2_helper import load_image
from transformers import TextIteratorStreamer
from threading import Thread


EXAMPLE_IMAGE = Path("examples_image1.jpg")
EXAMPLE_IMAGE_URL = "https://huggingface.co/OpenGVLab/InternVL2-2B/resolve/main/examples/image1.jpg"

if not EXAMPLE_IMAGE.exists():
    img_data = requests.get(EXAMPLE_IMAGE_URL).content
    with EXAMPLE_IMAGE.open("wb") as handler:
        handler.write(img_data)

pixel_values = load_image(EXAMPLE_IMAGE, max_num=12)

streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

generation_config = dict(max_new_tokens=100, do_sample=True, streamer=streamer)
question = "<image>\nPlease describe the image shortly."

display(PIL.Image.open(EXAMPLE_IMAGE))
print(f"User: {question}\n")
print("Assistant:")

thread = Thread(
    target=ov_model.chat,
    kwargs=dict(
        tokenizer=tokenizer,
        pixel_values=pixel_values,
        question=question,
        history=None,
        return_history=False,
        generation_config=generation_config,
    ),
)
thread.start()

generated_text = ""
# Loop through the streamer to get the new text as it is generated
for new_text in streamer:
    if new_text == ov_model.conv_template.sep:
        break
    generated_text += new_text
    print(new_text, end="", flush=True)  # Print each new chunk of generated text on the same line

../_images/internvl2-with-output_16_0.png

User: <image>
Please describe the image shortly.

Assistant:

Setting pad_token_id to eos_token_id:151645 for open-end generation.

The image shows a red panda lying on its side, partially wrapped in a wooden structure, possibly a container or log. The red panda appears to be looking at the camera with large, expressive eyes, displaying an endearing and lively appearance. The background consists of a portion of the red panda's habitat environment, which appears to be a tree and some greenery.

Interactive demo#

from gradio_helper import make_demo

demo = make_demo(ov_model, tokenizer)
try:
    demo.launch(debug=False, height=600)
except Exception:
    demo.launch(debug=False, share=True, height=600)
# if you are launching remotely, specify server_name and server_port
# demo.launch(server_name='your server name', server_port='server port in int')
# Read more in the docs: https://gradio.app/docs/

Running on local URL:  http://127.0.0.1:7860

To create a public link, set share=True in launch().