Instructions to use armand0e/Qwen3.6-35B-A3B-Fable-5-Distill with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use armand0e/Qwen3.6-35B-A3B-Fable-5-Distill with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="armand0e/Qwen3.6-35B-A3B-Fable-5-Distill")
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("armand0e/Qwen3.6-35B-A3B-Fable-5-Distill")
model = AutoModelForMultimodalLM.from_pretrained("armand0e/Qwen3.6-35B-A3B-Fable-5-Distill", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use armand0e/Qwen3.6-35B-A3B-Fable-5-Distill with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "armand0e/Qwen3.6-35B-A3B-Fable-5-Distill"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "armand0e/Qwen3.6-35B-A3B-Fable-5-Distill",
		"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/armand0e/Qwen3.6-35B-A3B-Fable-5-Distill

SGLang

How to use armand0e/Qwen3.6-35B-A3B-Fable-5-Distill 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 "armand0e/Qwen3.6-35B-A3B-Fable-5-Distill" \
    --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": "armand0e/Qwen3.6-35B-A3B-Fable-5-Distill",
		"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 "armand0e/Qwen3.6-35B-A3B-Fable-5-Distill" \
        --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": "armand0e/Qwen3.6-35B-A3B-Fable-5-Distill",
		"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
Docker Model Runner
How to use armand0e/Qwen3.6-35B-A3B-Fable-5-Distill with Docker Model Runner:
```
docker model run hf.co/armand0e/Qwen3.6-35B-A3B-Fable-5-Distill
```

Original tool calling issue persists

by weisunding - opened Jul 28

Discussion

weisunding

Jul 28

The original Qwen3.6 tool calling issue persists. (output tool calling as plain text, or not closing xml tag, ...)

armand0e

Owner Jul 28

Yea this seems to be a chat template level issue more than anything. It may help to use the froggeric template for Qwen3.6

armand0e

Owner Jul 28

Same with your earlier infinite thinking loop discussion (now deleted)

weisunding

Jul 29

OK, I will try a chat template, thanks!

The infinite loop of thinking

armand0e

Owner Jul 29

Wow I've never encountered this issue at all.. Can i asked more details on how you're serving the model and what sampling params you're using?

weisunding

Aug 1

I tried muliple chat templates, current used:
https://gist.github.com/jscott3201/e4b155885cc68c038d6ac8909a3bd9fe

My recipe

#!/bin/bash

NAME=qwen3.6-35b
PORT=3635

sudo docker stop $NAME > /dev/null 2>&1
sudo docker rm $NAME > /dev/null 2>&1

sleep 5

VLLM=vllm/vllm-openai:latest

sudo docker run \
    -d \
    --restart always \
    --runtime nvidia \
    --ipc=host \
    --gpus '"device=2,3"' \
    --name $NAME \
    -e HF_HUB_OFFLINE=1 \
    -e HF_ENDPOINT=https://hf-mirror.com \
    -e NCCL_P2P_DISABLE=1 \
    -e LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu \
    -v /data/model/huggingface:/root/.cache/huggingface \
    -v /data/model/modelscope:/root/.cache/modelscope \
    -v $(pwd)/chat:/chat \
    -p 127.0.0.1:$PORT:8000 \
    $VLLM \
    --model armand0e/Qwen3.6-35B-A3B-Fable-5-Distill \
    --served-model-name beast \
    --tensor-parallel-size 2 \
    --enable-auto-tool-choice \
    --reasoning-parser qwen3 \
    --tool-call-parser qwen3_coder \
    --chat-template /chat/chat_template.jinja \
    --default-chat-template-kwargs '{"enable_thinking": true}' \
    --speculative-config '{"method":"mtp","num_speculative_tokens":3}' \
    --gpu-memory-utilization 0.98 \
    --kv-cache-dtype fp8 \
    --max-model-len 204800 \
    --max-num-batched-tokens 8192 \
    --disable-custom-all-reduce \
    --enable-prefix-caching \
    --enable-chunked-prefill \
    --trust-remote-code

armand0e

Owner Aug 3

Interesting. Let me look into this more and get back to you. You said this is a known vllm issue?

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment