MiniCPM-V-2.6-OpenVINO-INT8

MiniCPM-V-2.6-OpenVINO-INT8 (Silver Series)

This repository contains the Silver Series optimized OpenVINO™ IR version of MiniCPM-V-2.6, quantized to INT8 precision using NNCF. This high-fidelity release offers elite-tier performance for multi-image reasoning and high-resolution visual analysis.

---

🐍 Python Inference (Optimum-Intel)

To run this Silver Series vision engine locally:

from optimum.intel import OVModelForVisualCausalLM
from transformers import AutoProcessor
from PIL import Image

model_id = "CelesteImperia/MiniCPM-V-2.6-OpenVINO-INT8"
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
model = OVModelForVisualCausalLM.from_pretrained(model_id, trust_remote_code=True)

image = Image.open("path/to/your/image.jpg")
prompt = "Perform a detailed audit of this visual scene."

inputs = processor(text=[prompt], images=[image], return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=256)
print(processor.decode(outputs[0], skip_special_tokens=True))

---

💻 For C# / .NET Users (OpenVINO.GenAI)

The MiniCPM-V-2.6 architecture is natively supported via the VLMPipeline in the OpenVINO.GenAI framework, enabling high-precision visual automation in industrial C# environments.

using OpenVino.GenAI;

// 1. Initialize the Visual-LLM Pipeline
var device = "CPU"; // Leverage "GPU" for accelerated dual-GPU inference
using var pipe = new VLMPipeline("path/to/minicpm-v-int8-model", device);

// 2. Load Visual Input
var image = OpenVino.GenAI.Utils.LoadImage("industrial_capture.jpg");
var prompt = "What are the specific technical identifiers visible on this component?";

// 3. Multimodal Execution
var result = pipe.Generate(prompt, image);
Console.WriteLine(result.Texts[0]);

---

🏗️ Technical Details

• Optimization Tool: NNCF (Neural Network Compression Framework)
• Quantization: INT8 Symmetric (Per-channel)
• Architecture: SigLip-400M + Qwen2-7B (8B Total)
• Series: Silver (High Fidelity)
• Workstation Validation: Dual-GPU (RTX 3090 + RTX A4000)

---

☕ Support the Forge

Maintaining the production line for high-fidelity multimodal models requires significant hardware and electrical resources. If these tools power your industrial projects or automation research, please consider supporting our development:

Scan to support via UPI (India Only):

<img src="https://huggingface.co/datasets/CelesteImperia/Assets/resolve/main/QrCode.jpeg" width="200">

---

📜 License

This model is released under the Apache 2.0 License.

---

Connect with the architect: Abhishek Jaiswal on LinkedIn