README.md

---
license: apache-2.0
language:
- en
library_name: transformers
pipeline_tag: text-generation
base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
tags:
- llm
- chatbot
- text-generation
- causal-lm
- harness
- algosciencelab
- vlsi
- coding
- reasoning
- pytorch
widget:
  - example_title: Python Example
    messages:
      - role: system
        content: You are Harness, an AI assistant created by Algo Science Lab.
      - role: user
        content: Write a Python function that calculates the factorial of a number.
---

<div align="center">

# Harness-1B

### Developed by Algo Science Lab

**Author:** Shahrear Hossain

**Hugging Face:** https://huggingface.co/algoscienceacademy

</div>

---

# Harness-1B

Harness-1B is a lightweight large language model developed by **Algo Science Lab** for instruction following, coding assistance, reasoning, mathematics, electronics, semiconductor engineering, VLSI design, and general conversational AI.

The model contains approximately **1 billion parameters**, making it suitable for local deployment while providing strong performance for everyday AI tasks.

Harness-1B has been fine-tuned to provide accurate, helpful, and concise responses across a wide variety of technical and general domains.

---

# Features

- General conversation
- Code generation
- Python programming
- C programming
- C++ programming
- Rust programming
- Java programming
- JavaScript
- TypeScript
- Verilog HDL
- SystemVerilog
- VHDL
- RTL Design
- FPGA Development
- ASIC Design
- CMOS Digital Design
- Semiconductor Engineering
- VLSI Design
- Mathematics
- Electronics
- Physics
- Problem Solving
- Technical Documentation
- AI Research Assistance

---

# Model Details

| Property | Value |
|----------|-------|
| Model Name | Harness-1B |
| Organization | Algo Science Lab |
| Hugging Face Username | algoscienceacademy |
| Parameters | ~1 Billion |
| Architecture | Llama-based |
| Model Type | Causal Language Model |
| Context Length | 2048 Tokens *(or your trained context size)* |
| Precision | FP16 / BF16 / GGUF |
| Framework | PyTorch |
| License | Apache-2.0 |

---

# Intended Uses

Harness-1B is designed for:

- AI Chatbots
- Programming Assistant
- Educational Applications
- Research
- Embedded AI
- Local AI Deployment
- Engineering Assistance
- Electronics Design
- FPGA Development
- ASIC/VLSI Workflow
- RTL Development
- Automation
- Documentation

---

# Example

```python
from transformers import pipeline

pipe = pipeline(
    "text-generation",
    model="algoscienceacademy/Harness-1B",
    device_map="auto"
)

messages = [
    {
        "role": "system",
        "content": "You are Harness, an AI assistant created by Algo Science Lab."
    },
    {
        "role": "user",
        "content": "Write a Python program to print Fibonacci numbers."
    }
]

prompt = pipe.tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

output = pipe(
    prompt,
    max_new_tokens=256,
    temperature=0.7,
)

print(output[0]["generated_text"])
```

---

# GGUF Usage

Harness-1B is also available in GGUF format for:

- llama.cpp
- LM Studio
- Jan
- Open WebUI
- KoboldCpp
- Ollama (after conversion)

Example:

```bash
./main \
-m Harness-1B-Q4_K_M.gguf \
-p "Explain CMOS Inverter."
```

---

# Training

Harness-1B is trained and fine-tuned using open-source datasets and instruction-following techniques.

Possible data sources include:

- SlimPajama
- StarCoderData
- UltraChat
- UltraFeedback

Additional custom datasets may have been used during supervised fine-tuning.

---

# Capabilities

Harness-1B can:

- Answer questions
- Explain concepts
- Generate code
- Debug code
- Write documentation
- Solve mathematics
- Explain algorithms
- Assist with VLSI
- Help with FPGA design
- Generate Verilog
- Generate SystemVerilog
- Produce technical reports

---

# Limitations

Harness-1B may:

- Produce incorrect information.
- Generate outdated knowledge.
- Make reasoning mistakes.
- Require verification for safety-critical applications.
- Require human review for production environments.

---

# Hardware Requirements

Recommended:

- 8 GB RAM (Q4 GGUF)
- 12 GB RAM (Q6 GGUF)
- 16 GB RAM (FP16)
- CUDA GPU recommended but optional

---

# Citation

```bibtex
@misc{Harness1B,
    title={Harness-1B},
    author={Algo Science Lab},
    year={2026},
    publisher={Hugging Face},
    howpublished={https://huggingface.co/algoscienceacademy/Harness-1B}
}
```

---

# License

Apache License 2.0

---

# Acknowledgements

Harness-1B builds upon open-source language model research and would not be possible without the work of the open-source AI community, including:

- Meta AI (Llama Architecture)
- TinyLlama Project
- Hugging Face
- PyTorch
- Transformers
- llama.cpp

---

# Contact

**Organization:** Algo Science Lab

**Hugging Face:** https://huggingface.co/algoscienceacademy

---

Made with ❤️ by Algo Science Lab.