113 lines
3.9 KiB
Markdown
113 lines
3.9 KiB
Markdown
---
|
||
license: apache-2.0
|
||
language:
|
||
- en
|
||
base_model:
|
||
- prithivMLmods/Kepler-Qwen3-4B-Super-Thinking
|
||
pipeline_tag: text-generation
|
||
library_name: transformers
|
||
tags:
|
||
- trl
|
||
- pytorch
|
||
- text-generation-inference
|
||
- thinking
|
||
- gpt_oss
|
||
- math
|
||
- code
|
||
- smoothing
|
||
- agent
|
||
---
|
||
|
||
|
||
|
||
# **Gliese-4B-OSS-0410**
|
||
|
||
> **Gliese-4B-OSS-0410** is a reasoning-focused model fine-tuned on **Qwen-4B** for enhanced **reasoning** and **polished token probability distributions**, delivering balanced **multilingual generation** across mathematics and general-purpose reasoning tasks.
|
||
> The model is fine-tuned on curated **GPT-OSS synthetic dataset entries**, improving its ability to handle structured reasoning, probabilistic inference, and multilingual tasks with precision.
|
||
|
||
> [!note]
|
||
> GGUF: [https://huggingface.co/prithivMLmods/Gliese-4B-OSS-0410-GGUF](https://huggingface.co/prithivMLmods/Gliese-4B-OSS-0410-GGUF)
|
||
|
||
---
|
||
|
||
## Key Features
|
||
|
||
1. **Enhanced Reasoning Precision**
|
||
Refined token probability distributions improve reasoning quality and ensure balanced, context-aware outputs.
|
||
|
||
2. **Event Simulation and Logical Analysis**
|
||
Capable of modeling random events, probability-driven reasoning, and structured decision-making with strong logical consistency.
|
||
|
||
3. **Multilingual Mathematical and General-Purpose Problem Solving**
|
||
Delivers robust performance in **mathematics**, **probability**, and **structured multilingual tasks**, enabling broad applicability in research and education.
|
||
|
||
4. **Hybrid Symbolic–Probabilistic Thinking**
|
||
Combines structured logic, probabilistic inference, and reasoning fluency to improve performance on uncertainty-driven tasks.
|
||
|
||
5. **Structured Output Generation**
|
||
Generates well-formatted outputs in **LaTeX**, **Markdown**, **JSON**, **CSV**, and **YAML**, supporting technical workflows and data-oriented research.
|
||
|
||
6. **Optimized Lightweight Footprint**
|
||
With **4B parameters**, it runs efficiently on **mid-range GPUs**, **offline clusters**, and **edge devices** without compromising reasoning performance.
|
||
|
||
---
|
||
|
||
## Quickstart with Transformers
|
||
|
||
```python
|
||
from transformers import AutoModelForCausalLM, AutoTokenizer
|
||
|
||
model_name = "prithivMLmods/Gliese-4B-OSS-0410"
|
||
|
||
model = AutoModelForCausalLM.from_pretrained(
|
||
model_name,
|
||
torch_dtype="auto",
|
||
device_map="auto"
|
||
)
|
||
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
||
|
||
prompt = "Simulate the probability of rolling two dice and getting a sum greater than 9. Show the reasoning."
|
||
|
||
messages = [
|
||
{"role": "system", "content": "You are a reasoning tutor skilled in probability, logic, and multilingual problem-solving."},
|
||
{"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
|
||
|
||
* Balanced multilingual reasoning and probability modeling
|
||
* Event simulation, uncertainty analysis, and structured problem solving
|
||
* Educational and research-focused reasoning tasks
|
||
* Deployment in mid-resource environments with efficient inference
|
||
* Structured technical content and data format generation
|
||
|
||
---
|
||
|
||
## Limitations
|
||
|
||
* Primarily focused on reasoning and mathematics; less suited for creative writing
|
||
* Despite its 4B size, extremely complex multi-hop reasoning tasks may remain challenging
|
||
* Prioritizes structured reasoning and probabilistic accuracy over conversational tone
|
||
* May produce inconsistent results with **very long contexts** or **cross-domain multi-document inputs** |