35 lines
1.3 KiB
Markdown
35 lines
1.3 KiB
Markdown
|
---
|
||
|
|
license: apache-2.0
|
||
|
|
base_model: Qwen/Qwen3-1.7B-Instruct
|
||
|
|
tags:
|
||
|
|
- agentic
|
||
|
|
- reasoning
|
||
|
|
- tool-use
|
||
|
|
- safetensors
|
||
|
|
- unsloth
|
||
|
|
- cpu-optimized
|
||
|
|
library_name: transformers
|
||
|
|
pipeline_tag: text-generation
|
||
|
|
---
|
||
|
|
|
||
|
|
# 🤖 Agentic AI Suite (Qwen3-1.7B-Reasoning)
|
||
|
|
|
||
|
|
This model is a **Small Reasoning Model (SRM)** fine-tuned specifically for the "Goldilocks Zone" of AI deployment: **Powerful enough** to handle PhD-level research tasks and **fast enough** to run on a standard Hugging Face Free Tier CPU.
|
||
|
|
|
||
|
|
By utilizing the **Qwen3-1.7B** architecture, this model achieves high-logic reasoning while maintaining a tiny memory footprint (~1.1GB in 4-bit).
|
||
|
|
|
||
|
|
## 🌟 Features
|
||
|
|
- **Native Tool Calling**: Trained to output structured `Action` calls for `web_search` and `calculator`.
|
||
|
|
- **ReAct Framework**: Uses a "Thought -> Action -> Observation -> Final Answer" loop.
|
||
|
|
- **Safetensors Format**: Merged 4-bit weights for instant loading and high-speed CPU inference via `transformers`.
|
||
|
|
- **Zero-Latency Logic**: Optimized to respond in under 10 seconds on a 2-vCPU environment.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## 🛠️ How to Use (Agentic Implementation)
|
||
|
|
|
||
|
|
To use this model as a true agent, your code should intercept the `Action:` text and execute the corresponding Python function.
|
||
|
|
|
||
|
|
### 1. Requirements
|
||
|
|
```bash
|
||
|
|
pip install transformers torch accelerate bitsandbytes
|