Create README.md

README.md

@@ -0,0 +1,113 @@
+---
+license: apache-2.0
+language:
+- en
+library_name: openvino
+tags:
+- openvino
+- qwen2
+- civil-engineering
+- structural-engineering
+- building-engineering
+- text-generation
+- conversational
+datasets:
+- Irfanuruchi/buildeng-v8-1.5b
+base_model:
+- Qwen/Qwen2.5-1.5B-Instruct
+---
+
+# BuildEng 1.5B OpenVINO (INT8)
+
+Intel optimized version of BuildEng 1.5B.
+
+This INT8 version keeps more precision than INT4 while still staying lightweight and practical for OpenVINO inference on Intel hardware.
+
+## What can it help with?
+
+- Building defects and pathology
+- Reinforced concrete and masonry
+- Foundation and settlement issues
+- Structural inspection guidance
+- Construction best practices
+- Civil and building engineering Q&A
+
+## Important
+
+BuildEng is an AI assistant. It should not replace professional engineering judgment, real site inspections, licensed structural assessments, or local building codes.
+
+Use it as support, not as final authority.
+
+---
+
+## Base Model
+
+- Qwen2.5-1.5B-Instruct
+
+## Fine-tuned by
+
+- Irfan Uruçi
+
+---
+
+## Why this version?
+
+- Higher precision than INT4
+- Still small and efficient
+- Good choice when quality matters more than smallest size
+
+---
+
+## Installation
+
+```bash
+pip install "optimum-intel[openvino]" openvino transformers
+```
+
+## Usage
+
+```python
+from optimum.intel.openvino import OVModelForCausalLM
+from transformers import AutoTokenizer
+
+model_id = "Irfanuruchi/qwen2.5-1.5b-buildeng-openvino-int8"
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = OVModelForCausalLM.from_pretrained(model_id, device="CPU", compile=True)
+
+messages = [
+    {"role": "system", "content": "You are BuildEng. Give concise, inspection-first building engineering guidance."},
+    {"role": "user", "content": "A homeowner reports diagonal cracks near a window corner. What should be inspected first?"}
+]
+
+prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+inputs = tokenizer(prompt, return_tensors="pt")
+
+output = model.generate(**inputs, max_new_tokens=160, do_sample=False)
+new_tokens = output[0][inputs["input_ids"].shape[-1]:]
+
+print(tokenizer.decode(new_tokens, skip_special_tokens=True))
+```
+
+## Devices
+
+- CPU
+- GPU
+- NPU, where supported by OpenVINO and the system
+
+---
+
+## Other Formats
+
+| Format | Repository |
+|---|---|
+| Hugging Face | `Irfanuruchi/qwen2.5-1.5b-buildeng` |
+| GGUF | `Irfanuruchi/qwen2.5-1.5b-buildeng-GGUF-Q4_K_M` |
+| MLX 4-bit | `Irfanuruchi/qwen2.5-1.5b-buildeng-mlx-4bit` |
+| MLX 8-bit | `Irfanuruchi/qwen2.5-1.5b-buildeng-mlx-8bit` |
+| OpenVINO INT4 | `Irfanuruchi/qwen2.5-1.5b-buildeng-openvino-int4` |
+
+---
+
+Built by **Irfan Uruçi**  
+Computer Engineer