146 lines
5.8 KiB
Markdown
146 lines
5.8 KiB
Markdown
---
|
|
license: apache-2.0
|
|
datasets:
|
|
- Salesforce/xlam-function-calling-60k
|
|
language:
|
|
- en
|
|
base_model:
|
|
- Qwen/Qwen3-0.6B
|
|
pipeline_tag: text-generation
|
|
library_name: transformers
|
|
tags:
|
|
- trl
|
|
- text-generation-inference
|
|
- agent
|
|
- tool_calling
|
|
- fun()
|
|
---
|
|
|
|
|
|
|
|
# **Gliese-Query_Tool-0.6B**
|
|
|
|
> **Gliese-Query_Tool-0.6B** is a **function-calling and query-oriented reasoning model** fine-tuned from **Qwen3-0.6B** using **Salesforce/xlam-function-calling-60k**, designed for **tool orchestration**, **structured query resolution**, and **operation chaining** across diverse tasks.
|
|
> It excels in **dynamic function execution**, **structured reasoning pipelines**, and **multi-tool decision workflows**, making it a powerful lightweight solution for **developers**, **tooling platforms**, and **automation systems**.
|
|
|
|
> [!note]
|
|
> GGUF: [https://huggingface.co/prithivMLmods/Gliese-Query_Tool-0.6B-GGUF](https://huggingface.co/prithivMLmods/Gliese-Query_Tool-0.6B-GGUF)
|
|
|
|
---
|
|
|
|
## **Key Features**
|
|
|
|
1. **Function-Calling Focused Reasoning**
|
|
Fine-tuned with **Salesforce/xlam-function-calling-60k**, enabling precise function selection, argument formatting, and multi-step tool invocation in complex workflows.
|
|
|
|
2. **Query-Oriented Workflow Design**
|
|
Built to parse, interpret, and resolve complex queries by selecting and chaining the most relevant functions or tools for the task.
|
|
|
|
3. **Tool-Orchestration & Automation**
|
|
Handles structured tool calls, dynamic function dispatch, and hybrid reasoning to power intelligent automation, API orchestration, and backend agent pipelines.
|
|
|
|
4. **Structured Multi-Format Output**
|
|
Outputs formatted responses in **JSON**, **YAML**, **Markdown**, and **structured argument objects** — ideal for direct integration into software pipelines and agentic systems.
|
|
|
|
5. **Lightweight, Deployment-Ready Core**
|
|
Compact **0.6B parameter** size optimized for **edge deployments**, **on-device inference**, and **fast cold-starts** while maintaining strong reasoning and function-call accuracy.
|
|
|
|
---
|
|
|
|
## **sample inference.**
|
|
|
|
```
|
|
Solve 2**2
|
|
```
|
|
|
|
```py
|
|
[{"name": "power", "description": "Calculates the power of a number with a specified exponent.",
|
|
"parameters": {"number": {"description": "The base for which the power is calculated.", "type": "int"},
|
|
"exponent": {"description": "The exponent to which the number should be raised.", "type": "int"}}}]
|
|
```
|
|
|
|
```
|
|
solve for 'x' in the equation 2x + 5 = 11?
|
|
```
|
|
|
|
```py
|
|
[{"name": "solving_equation", "description": "Solves a linear equation for a variable.",
|
|
"parameters": {"equation": {"description": "The equation to solve. The format is 'a*x + b = c'.
|
|
For example, '5x + 2 = 10' or '3x - 7 = 1'.", "type": "str"}, "operation": {"description": "The operation (add, sub, etc.) to perform the solving.",
|
|
"type": "str, optional"}, "variable": {"description": "The variable to solve for. Defaults to 'x' if not provided.", "default": "x"}}}]
|
|
```
|
|
|
|
```
|
|
What is the volume of a sphere with a radius of 6 cm?
|
|
```
|
|
|
|
```py
|
|
[{"name": "volume_of_sphere", "description": "Calculates the volume of a sphere given its radius using the formula (4/3)πr³.",
|
|
"parameters": {"radius": {"description": "The radius of the sphere.", "type": "int"}}}]
|
|
```
|
|
|
|
```
|
|
In an examination 80% of the candidates passed in Urdu and 85% in Hindi, while 75% passed in both . If 45 candidates failed in both. Then the total number of candidates was ?
|
|
```
|
|
|
|
```py
|
|
[{"name": "passing_percentage", "description": "Calculates the passing percentage for an exam given the percentage of students who passed each subject, and the intersection percentage of passing subjects.",
|
|
"parameters": {"subject1_percent": {"description": "Percentage of students who passed the first subject (e.g., 85% if Hindi).", "type": "int"},
|
|
"subject2_percent": {"description": "Percentage of students who passed the second subject (e.g., 80% if Urdu).", "type": "int"}, "passed_both_percent": {"description": "Percentage of students who passed both subjects.", "type": "int"}}}]
|
|
```
|
|
|
|
## **Quickstart with Transformers**
|
|
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
model_name = "prithivMLmods/Gliese-Query_Tool-0.6B"
|
|
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|
model_name,
|
|
torch_dtype="auto",
|
|
device_map="auto"
|
|
)
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|
|
|
prompt = "Call the right function to fetch weather data for Paris and format the output as JSON."
|
|
|
|
messages = [
|
|
{"role": "system", "content": "You are a query tool model skilled in function-calling, API orchestration, and structured query resolution."},
|
|
{"role": "user", "content": prompt}
|
|
]
|
|
|
|
text = tokenizer.apply_chat_template(
|
|
messages,
|
|
tokenize=False,
|
|
add_generation_prompt=True
|
|
)
|
|
|
|
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
|
|
|
|
generated_ids = model.generate(
|
|
**model_inputs,
|
|
max_new_tokens=512
|
|
)
|
|
generated_ids = [
|
|
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
|
|
]
|
|
|
|
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
|
print(response)
|
|
```
|
|
|
|
## **Intended Use**
|
|
|
|
* Intelligent function-calling and multi-step query solving
|
|
* API orchestration, agent tool selection, and dynamic workflows
|
|
* Structured data generation and backend reasoning integration
|
|
* Lightweight agentic systems and on-device automation
|
|
* Developer-focused query resolution and toolchain automation
|
|
|
|
## **Limitations**
|
|
|
|
* Focused on **function-calling and structured tasks** — not suited for open-ended dialogue or creative writing
|
|
* Small model size means very complex reasoning chains may require external planning agents
|
|
* Optimized for structured tool workflows — conversational tone and narrative depth are secondary
|
|
* Long-context multi-tool planning beyond several steps may reduce reliability |