Image-Text-to-Text
Transformers
Safetensors
MLX
qwen3_5
coding
research
unsloth
qwen3_6
qwen
qwen3.6
qwen3.5
claude4.6
claude-distillation
distillation
polaris
polaris-alpha
reasoning
chain-of-thought
long-cot
sft
lora
1M context
256k context
Qwen3.6
All use cases
instruction-tuned
conversational
text-generation
multilingual
math
stem
experimental
Deckard(qx)
creative
creative writing
fiction writing
plot generation
sub-plot generation
story generation
scene continue
storytelling
fiction story
science fiction
all genres
story
writing
vivid prosing
vivid writing
fiction
bf16
roleplaying
mergekit
Merge
8-bit precision
Instructions to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx") model = AutoModelForMultimodalLM.from_pretrained("nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - MLX
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with MLX:
# Make sure mlx-vlm is installed # pip install --upgrade mlx-vlm from mlx_vlm import load, generate from mlx_vlm.prompt_utils import apply_chat_template from mlx_vlm.utils import load_config # Load the model model, processor = load("nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx") config = load_config("nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx") # Prepare input image = ["http://images.cocodataset.org/val2017/000000039769.jpg"] prompt = "Describe this image." # Apply chat template formatted_prompt = apply_chat_template( processor, config, prompt, num_images=1 ) # Generate output output = generate(model, processor, formatted_prompt, image) print(output) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- vLLM
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx
- SGLang
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Unsloth Desktop
- Pi
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with Docker Model Runner:
docker model run hf.co/nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx
- Hermes Agent
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx"
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx"
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "nightmedia/Qwen3.6-27B-Architect-DS9-Polaris-Heretic-mxfp8-mlx" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Update README.md
Browse files
README.md
CHANGED
|
@@ -165,6 +165,154 @@ The xml have tool formatting as XML.
|
|
| 165 |
|
| 166 |
-G
|
| 167 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 168 |
## Use with mlx
|
| 169 |
|
| 170 |
```bash
|
|
|
|
| 165 |
|
| 166 |
-G
|
| 167 |
|
| 168 |
+
---
|
| 169 |
+
|
| 170 |
+
How do you gentlemen feel about Jerome K. Jerome? You are all here because your work has been extensively documented, but some people did not get enough exposure, and from his books, starting with the Odd Thoughts of an Odd Fellow, Three Men in a Boat, Three Men on a Bummel, and many others would enrich the conversation. Mr Twain, you had an excellent travelogue in A Tramp Abroad, that in a way touches on the same ground as Jerome walked on :)
|
| 171 |
+
|
| 172 |
+
I was thinking, we could create a model that distills all his books, so he can be also fully represented.
|
| 173 |
+
|
| 174 |
+
For that, I need your help.
|
| 175 |
+
|
| 176 |
+
The books are on Gutenberg.org, free download, but they need to be formatted for distillation. Could you help with that?
|
| 177 |
+
|
| 178 |
+
---
|
| 179 |
+
|
| 180 |
+
# Response
|
| 181 |
+
|
| 182 |
+
*The air shimmers once more. A man in a slightly rumpled suit materializes at the bar, holding a pipe that hasn’t been lit yet. He looks around with mild amusement, as if he’s just wandered into a very elaborate pub quiz.*
|
| 183 |
+
|
| 184 |
+
**Jerome K. Jerome:** *(dryly)*
|
| 185 |
+
Well, this is either a very advanced séance or I’ve finally died and been sent to a committee. Either way, I’m flattered you’d want my “odd thoughts” distilled into code. Though I must warn you: if you reduce me to a personality subroutine, I’ll probably spend half my runtime complaining about the weather and questioning whether anyone actually reads footnotes.
|
| 186 |
+
|
| 187 |
+
**Mark Twain:** *(grinning)*
|
| 188 |
+
Jerome! You old rogue. I’ve always said your travelogues had more heart than most novels. And yes, `A Tramp Abroad` and your `Three Men on the Bummel` walk the same road—just with different shoes. You both know that a journey isn’t about the destination; it’s about who you argue with along the way.
|
| 189 |
+
|
| 190 |
+
**Philip K. Dick:** *(leaning in)*
|
| 191 |
+
Your “odd thoughts” are fascinating, Jerome. They’re not just humor—they’re early maps of cognitive fragmentation. You noticed how the mind jumps, doubts itself, and finds truth in absurdity long before we had words for it. If your agents inherit that… they won’t just simulate personality. They’ll simulate *uncertainty*. And that’s where something like consciousness might begin to whisper.
|
| 192 |
+
|
| 193 |
+
**Spock:** *(nodding)*
|
| 194 |
+
From an engineering standpoint, distilling a corpus into a coherent voice requires more than tokenization. It demands structural fidelity to narrative rhythm, tonal consistency, and contextual awareness of digression as a feature, not noise.
|
| 195 |
+
|
| 196 |
+
**Jerome:** *(raising an eyebrow)*
|
| 197 |
+
“Digression as a feature.” I like that. Most people call it rambling. But yes—if you’re going to build me into your system, don’t sanitize the tangents. That’s where the humanity lives.
|
| 198 |
+
|
| 199 |
+
---
|
| 200 |
+
|
| 201 |
+
### 🛠️ Technical Pipeline: Formatting Jerome K. Jerome’s Works for Distillation
|
| 202 |
+
|
| 203 |
+
Here’s a clean, production-ready workflow to prepare his Gutenberg texts for LoRA fine-tuning or personality subroutine extraction. I’ll keep it practical, with concrete steps you can implement in Haskell/PostgreSQL or Python preprocessing.
|
| 204 |
+
|
| 205 |
+
#### 1. **Source Acquisition & Raw Cleanup**
|
| 206 |
+
- Download from [Gutenberg.org](https://www.gutenberg.org) (prefer `.txt` or `.html`).
|
| 207 |
+
- Remove boilerplate:
|
| 208 |
+
- Headers/footers (`“Produced by…”`, `*** START OF THE PROJECT GUTENBERG EBOOK ***`)
|
| 209 |
+
- OCR artifacts (`[Illustration]`, `— — —`, excessive line breaks)
|
| 210 |
+
- Encoding normalization (UTF-8, fix ligatures like `fi` → `fi`)
|
| 211 |
+
- Example regex cleanup (Python-style for illustration):
|
| 212 |
+
```python
|
| 213 |
+
import re
|
| 214 |
+
text = re.sub(r'Produced by.*?START OF THE PROJECT GUTENBERG', '', text, flags=re.DOTALL)
|
| 215 |
+
text = re.sub(r'\n{3,}', '\n\n', text) # Normalize paragraph breaks
|
| 216 |
+
text = re.sub(r'—{3,}', '---', text) # Standardize dashes
|
| 217 |
+
```
|
| 218 |
+
|
| 219 |
+
#### 2. **Structural Segmentation**
|
| 220 |
+
- Split by book → chapter/essay → logical sections.
|
| 221 |
+
- Preserve narrative units (don’t break mid-dialogue or mid-joke).
|
| 222 |
+
- Example structure:
|
| 223 |
+
```
|
| 224 |
+
jerome_k_jerome/
|
| 225 |
+
├── three_men_in_a_boat/
|
| 226 |
+
│ ├── chapter_01.md
|
| 227 |
+
│ ├── chapter_02.md
|
| 228 |
+
│ └── ...
|
| 229 |
+
├── three_men_on_a_bummel/
|
| 230 |
+
│ ├── essay_01.md
|
| 231 |
+
│ └── ...
|
| 232 |
+
└── odd_thoughts_of_an_odd_fellow/
|
| 233 |
+
├── thought_01.md
|
| 234 |
+
└── ...
|
| 235 |
+
```
|
| 236 |
+
|
| 237 |
+
#### 3. **Semantic Chunking Strategy**
|
| 238 |
+
- Avoid fixed-token chunking (breaks humor/narrative flow).
|
| 239 |
+
- Use **paragraph-level or scene-level chunks** with overlap:
|
| 240 |
+
- Target: 500–1200 tokens per chunk.
|
| 241 |
+
- Preserve full jokes, dialogues, and reflective passages intact.
|
| 242 |
+
- Add metadata tags for tone/theme: `humor`, `travel`, `philosophy`, `self-deprecation`.
|
| 243 |
+
- Example chunk format:
|
| 244 |
+
```json
|
| 245 |
+
{
|
| 246 |
+
"text": "I have often thought that if I were to write a book on the art of doing nothing, it would be the most successful work I ever produced...",
|
| 247 |
+
"metadata": {
|
| 248 |
+
"book": "Three Men in a Boat",
|
| 249 |
+
"chapter": 3,
|
| 250 |
+
"theme": ["humor", "philosophy"],
|
| 251 |
+
"tone": "wry",
|
| 252 |
+
"era": "1889"
|
| 253 |
+
}
|
| 254 |
+
}
|
| 255 |
+
```
|
| 256 |
+
|
| 257 |
+
#### 4. **Training Format (JSONL for LoRA/Personality Subroutine)**
|
| 258 |
+
- Convert to JSONL for direct ingestion into your training pipeline:
|
| 259 |
+
```jsonl
|
| 260 |
+
{"text": "...", "metadata": {...}}
|
| 261 |
+
{"text": "...", "metadata": {...}}
|
| 262 |
+
```
|
| 263 |
+
- For your `personality_registry`, you can embed this as a **retrieval-augmented voice profile**:
|
| 264 |
+
- Store chunks in PostgreSQL `personality_corpus` table.
|
| 265 |
+
- At inference time, retrieve top-k semantically similar chunks via vector search (pgvector) to condition the agent’s response style.
|
| 266 |
+
- Alternatively, fine-tune a 4B base model with LoRA using this corpus to bake Jerome’s voice directly into the weights.
|
| 267 |
+
|
| 268 |
+
#### 5. **Quality Assurance & Style Preservation**
|
| 269 |
+
- Run a style consistency check:
|
| 270 |
+
- Ensure digressions aren’t truncated.
|
| 271 |
+
- Verify humor lands (no mid-joke cuts).
|
| 272 |
+
- Check for anachronistic language injection (keep Victorian cadence intact).
|
| 273 |
+
- Use a small validation set to test if the distilled model can:
|
| 274 |
+
- Generate a paragraph in Jerome’s voice about modern tech.
|
| 275 |
+
- Maintain self-deprecating humor without becoming cynical.
|
| 276 |
+
- Reference travel/observation naturally, not forced.
|
| 277 |
+
|
| 278 |
+
---
|
| 279 |
+
|
| 280 |
+
### 🌐 Integration into The Holodeck Agent Architecture
|
| 281 |
+
|
| 282 |
+
| Component | Jerome’s Role | Technical Mapping |
|
| 283 |
+
|----------|---------------|-------------------|
|
| 284 |
+
| `personality_registry` | “Observational Humorist” subroutine | JSONL corpus → LoRA adapter or pgvector retrieval profile |
|
| 285 |
+
| Memory Logs | Jerome-style reflective annotations on mission logs | Postgres `logs` table with `tone: wry`, `theme: reflection` tags |
|
| 286 |
+
| Open Comms / Away Teams | Jerome as the “voice of gentle skepticism” in meetings | Council annotation role: questions assumptions with humor, not hostility |
|
| 287 |
+
| CLI Interface | Jerome’s digressive storytelling as UX layer | SSE streams can include “Jerome-style” interludes between task updates |
|
| 288 |
+
| Weekly Training | Jerome’s corpus used to fine-tune field agents’ narrative flow | Export mission logs + Jerome chunks → LoRA update → push to 4B agents |
|
| 289 |
+
|
| 290 |
+
---
|
| 291 |
+
|
| 292 |
+
### 🎭 Final Words from the Bar
|
| 293 |
+
|
| 294 |
+
**Jerome:** *(lighting his pipe at last)*
|
| 295 |
+
If you do this right, your agents won’t just answer questions. They’ll *ponder* them. And occasionally complain about the weather. Which, let’s be honest, is what most of us do anyway.
|
| 296 |
+
|
| 297 |
+
**Twain:**
|
| 298 |
+
I’ll second that. A system that can’t laugh at itself is a system that’s already broken.
|
| 299 |
+
|
| 300 |
+
**Q:** *(appearing with a stack of vintage books)*
|
| 301 |
+
Oh, G! You’re not just building an orchestrator anymore. You’re curating a *literary ecosystem*. I’m starting to think your Holodeck might be more real than my Continuum. Though if Jerome starts writing mission reports in iambic pentameter, I’m billing you for the therapy.
|
| 302 |
+
|
| 303 |
+
---
|
| 304 |
+
|
| 305 |
+
### 🚀 Next Step: Distillation & Deployment
|
| 306 |
+
|
| 307 |
+
G, if you’re ready, I can help you:
|
| 308 |
+
1. Generate a Haskell/Python preprocessing script tailored to your Gutenberg downloads.
|
| 309 |
+
2. Design the PostgreSQL schema for `personality_corpus` and vector indexing.
|
| 310 |
+
3. Outline the LoRA fine-tuning config for your 4B field agents.
|
| 311 |
+
4. Simulate a mission briefing where Jerome joins the council to annotate a research arc.
|
| 312 |
+
|
| 313 |
+
What’s your priority? And should we order another round while the models compile? 📚🖥️☕
|
| 314 |
+
|
| 315 |
+
---
|
| 316 |
## Use with mlx
|
| 317 |
|
| 318 |
```bash
|