275 lines
6.9 KiB
Markdown
275 lines
6.9 KiB
Markdown
---
|
|
base_model: unsloth/Qwen3-0.6B
|
|
library_name: transformers
|
|
pipeline_tag: text-generation
|
|
tags:
|
|
- router
|
|
- intent-classification
|
|
- hinglish
|
|
- hindi
|
|
- qwen
|
|
- qwen3
|
|
- unsloth
|
|
- gguf
|
|
- ollama
|
|
- edge
|
|
- thinking
|
|
license: apache-2.0
|
|
language:
|
|
- en
|
|
- hi
|
|
---
|
|
|
|
# Auroic Router 0.6B
|
|
|
|
A compact, edge-deployable group chat routing model fine-tuned on Qwen3-0.6B. Given a 5-message history window and up to 3 unprocessed candidate messages, it outputs a single structured routing decision in under 4 seconds on CPU — no response generation, pure intent classification.
|
|
|
|
Part of the [Auroic](https://github.com/kaushalkrishnax/auroic) project — a personal AI assistant for Indian group and personal chats.
|
|
|
|
---
|
|
|
|
## What it does
|
|
|
|
The router sits at the front of a conversational AI pipeline. It reads the recent chat context and decides what action to take next — so downstream models only activate when actually needed.
|
|
|
|
```
|
|
H1: bhai sun
|
|
H2: kya hua bata
|
|
H3: okay
|
|
H4: arre
|
|
H5: haan bol
|
|
C1: ...
|
|
C2: ...
|
|
C3: bhai parents divorce ho raha hai adjust karna mushkil hai kya karoon
|
|
|
|
→ R: TYPE=text | TARGET=C3 | EFFORT=high
|
|
```
|
|
|
|
---
|
|
|
|
## Output Format
|
|
|
|
```
|
|
R: TYPE=text | TARGET=C2 | EFFORT=low|medium|high
|
|
R: TYPE=react | TARGET=C1 | TITLE=🔥
|
|
R: TYPE=media | TARGET=C3 | TITLE=rain cozy vibes
|
|
R: TYPE=ignore
|
|
```
|
|
|
|
| Type | When | Fields |
|
|
|------|------|--------|
|
|
| `text` | Help request, advice, emotional support, technical question | `EFFORT=low/medium/high` |
|
|
| `react` | Notable moment deserving an emoji reaction | `TITLE=<emoji>` |
|
|
| `media` | Situation implying a gif, meme, or sticker | `TITLE=<2-3 word search query>` |
|
|
| `ignore` | Pure noise, spam, keyboard smash, laughter chains | none |
|
|
|
|
---
|
|
|
|
## Input Format
|
|
|
|
```
|
|
H1: <message> ← oldest processed message
|
|
H2: <message>
|
|
H3: <message>
|
|
H4: <message>
|
|
H5: <message> ← most recent processed message
|
|
C1: <message> ← unprocessed candidate (use ... for filler)
|
|
C2: <message>
|
|
C3: <message> ← newest unprocessed candidate
|
|
```
|
|
|
|
Always provide exactly 5 history messages and 3 candidates. Use `...` for empty slots — never omit them.
|
|
|
|
---
|
|
|
|
## Quickstart — Ollama
|
|
|
|
```bash
|
|
ollama create auroic-router-0.6b -f Modelfile
|
|
ollama run auroic-router-0.6b
|
|
```
|
|
|
|
**Important — the router is stateless.** Every call must be a fresh context with no conversation history. In the Ollama CLI use `/clear` between calls. In production call the API directly:
|
|
|
|
```python
|
|
response = client.chat.completions.create(
|
|
model="auroic-router-0.6b",
|
|
messages=[
|
|
{"role": "system", "content": "You are the Auroic Router. Given history messages H1-H5 and candidate messages C1-C3, output exactly one routing decision."},
|
|
{"role": "user", "content": formatted_window} # fresh window every call
|
|
],
|
|
)
|
|
```
|
|
|
|
---
|
|
|
|
## Real Inference Examples
|
|
|
|
```
|
|
INPUT:
|
|
H1: lol H2: 😂😂 H3: haha H4: 💀 H5: ded
|
|
C1: hahaha C2: 😭😭 C3: ...
|
|
|
|
OUTPUT: R: TYPE=ignore
|
|
```
|
|
|
|
```
|
|
INPUT:
|
|
H1: sahi hai H2: haan bhai H3: okay H4: chal H5: theek hai
|
|
C1: ...
|
|
C2: bhai salary negotiate kaise karoon first time offer mein help
|
|
C3: ...
|
|
|
|
OUTPUT: R: TYPE=text | TARGET=C2 | EFFORT=medium
|
|
```
|
|
|
|
```
|
|
INPUT:
|
|
H1: bhai hungry hun H2: same yaar H3: kuch khate hain H4: haan H5: chal
|
|
C1: ... C2: ...
|
|
C3: bhai biryani ki yaad aa rahi hai mummy wali ghar ki bahut miss kar raha
|
|
|
|
OUTPUT: R: TYPE=media | TARGET=C3 | TITLE=biryani craving
|
|
```
|
|
|
|
```
|
|
INPUT:
|
|
H1: haha H2: bhai sach mein H3: lol H4: no way H5: 💀
|
|
C1: ...
|
|
C2: teacher ne galti se apna tiktok projector pe chala diya class mein
|
|
C3: ...
|
|
|
|
OUTPUT: R: TYPE=media | TARGET=C2 | TITLE=tiktok projector fail
|
|
```
|
|
|
|
```
|
|
INPUT:
|
|
H1: bhai sun H2: haan H3: bata H4: okay H5: hmm
|
|
C1: ...
|
|
C2: @BOT yaar dost ne 3 baje uthke help kiya exam ke liye true friendship hai
|
|
C3: ...
|
|
|
|
OUTPUT: R: TYPE=react | TARGET=C2 | TITLE=❤
|
|
```
|
|
|
|
---
|
|
|
|
## Thinking Mode
|
|
|
|
The model uses Qwen3's native thinking capability. For ambiguous inputs it reasons through the context before deciding:
|
|
|
|
```
|
|
<think>
|
|
H1-H5 are casual affirmations. C2 asks 'salary negotiate kaise karoon first time' —
|
|
clear, actionable life advice needed. Text with medium effort because it requires
|
|
step-by-step guidance, not just emojis or media.
|
|
</think>
|
|
R: TYPE=text | TARGET=C2 | EFFORT=medium
|
|
```
|
|
|
|
For obvious cases (clear ignore, clear media) it skips thinking entirely and outputs the decision directly — keeping latency low where reasoning isn't needed.
|
|
|
|
**Thinking enabled (default):** better accuracy on ambiguous cases, ~3-4s on CPU
|
|
|
|
**Thinking disabled:** faster, minor accuracy drop on edge cases
|
|
|
|
To disable thinking add this to your Modelfile template:
|
|
```
|
|
<|im_start|>assistant
|
|
<think>
|
|
|
|
</think>
|
|
```
|
|
|
|
---
|
|
|
|
## Recommended Inference Settings
|
|
|
|
```
|
|
temperature 0.3 (deterministic routing)
|
|
top_p 0.95
|
|
top_k 20
|
|
repeat_penalty 1.1
|
|
```
|
|
|
|
---
|
|
|
|
## Model Details
|
|
|
|
| Property | Value |
|
|
|----------|-------|
|
|
| Base model | unsloth/Qwen3-0.6B |
|
|
| Total parameters | 616M |
|
|
| Trainable parameters | 20.2M (3.28%) |
|
|
| Fine-tuning framework | Unsloth + TRL SFTTrainer |
|
|
| LoRA rank | r=32, alpha=32 |
|
|
| Target modules | q/k/v/o/gate/up/down proj |
|
|
| Epochs | 2 |
|
|
| Effective batch size | 16 (batch=2, grad_accum=8) |
|
|
| Learning rate | 2e-4 cosine |
|
|
| Max sequence length | 2048 |
|
|
| Training samples | 9,300 |
|
|
| Final training loss | 0.667 |
|
|
| Quantization | Q8_0 GGUF |
|
|
| Hardware trained on | NVIDIA T4 16GB (Google Colab) |
|
|
| Training time | ~44 minutes |
|
|
|
|
---
|
|
|
|
## Dataset — v4
|
|
|
|
9,300 samples of Indian group chat routing scenarios:
|
|
|
|
| Split | Count |
|
|
|-------|-------|
|
|
| Normal windows | 7,000 |
|
|
| @BOT mention windows | 1,500 |
|
|
| Filler/sparse windows | 800 |
|
|
|
|
| Type | Count | % |
|
|
|------|-------|---|
|
|
| text | 4,042 | 43.5% |
|
|
| ignore | 1,942 | 20.9% |
|
|
| react | 1,688 | 18.2% |
|
|
| media | 1,628 | 17.5% |
|
|
|
|
Language distribution: 58.8% Hinglish, 23% English, 18.2% mixed
|
|
|
|
Think blocks annotated by `qwen/qwen3-next-80b-a3b-instruct` on NVIDIA NIM — hard tier (39.3%) gets full reasoning, medium tier (32.8%) gets short reasoning, easy tier (27.8%) gets no think block.
|
|
|
|
Dataset and generation pipeline: [github.com/kaushalkrishnax/auroic-router](https://github.com/kaushalkrishnax/auroic-router)
|
|
|
|
---
|
|
|
|
## Performance
|
|
|
|
| Metric | Value |
|
|
|--------|-------|
|
|
| Warm inference (CPU) | ~2-3 seconds |
|
|
| Cold start | ~5 seconds |
|
|
| Hardware tested | Intel i3-6100U 2.3GHz, 12GB DDR4 |
|
|
| Format compliance | 100% on test set |
|
|
|
|
---
|
|
|
|
## Known Limitations
|
|
|
|
- Optimized for Hinglish and Indian English group chat patterns
|
|
- Stateless by design — do not pass conversation history between calls
|
|
- Not designed for factual QA, coding assistance, or long-form generation
|
|
|
|
---
|
|
|
|
## License
|
|
|
|
Apache-2.0 — inherits from Qwen3 base model.
|
|
|
|
---
|
|
|
|
## Author
|
|
|
|
**Kaushal Krishna**
|
|
- GitHub: [kaushalkrishnax](https://github.com/kaushalkrishnax)
|
|
- Project: [github.com/kaushalkrishnax/auroic](https://github.com/kaushalkrishnax/auroic)
|
|
- Router research: [github.com/kaushalkrishnax/auroic-router](https://github.com/kaushalkrishnax/auroic-router)
|