Instructions to use cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic with Transformers:

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

pipe = pipeline("image-text-to-text", model="cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic")
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("cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic")
model = AutoModelForMultimodalLM.from_pretrained("cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic", 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 cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic",
		"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/cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic

SGLang

How to use cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic 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 "cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic" \
    --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": "cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic",
		"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 "cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic" \
        --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": "cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic",
		"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"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic with Docker Model Runner:
```
docker model run hf.co/cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic
```

cloud19 commited on Aug 14

Commit

5fde52b

verified ·

1 Parent(s): e50a71c

Model card in English

Browse files

Files changed (1) hide show

README.md +29 -16

README.md CHANGED Viewed

@@ -4,6 +4,7 @@ license_name: apache-2.0
 library_name: transformers
 pipeline_tag: image-text-to-text
 base_model: joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive
 tags:
 - fp8
 - llm-compressor
@@ -20,22 +21,26 @@ language:
 # JoyFox-Qwen3.6-35B-A3B-RP-Aggressive — FP8 Dynamic
-FP8_DYNAMIC-квант модели
 [joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive](https://huggingface.co/joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive)
-(архитектура `Qwen3_5MoeForConditionalGeneration`, база — `Qwen/Qwen3.6-35B-A3B`).
-* **Источник:** `joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive`
-* **Лицензия:** apache-2.0 (наследуется от источника и базовой модели)
-* **Инструмент:** [llm-compressor](https://github.com/vllm-project/llm-compressor), data-free one-shot
-* **Схема:** веса FP8 per-channel (статические), активации FP8 per-token (динамические)
-## Что не квантовалось
-В BF16 оставлены: `lm_head`, эмбеддинги, вижн-башня (`model.visual.*`),
-слои линейного внимания Gated DeltaNet (`linear_attn.*`), роутеры MoE (`mlp.gate`),
-gate шаренного эксперта (`shared_expert_gate`) и MTP-голова (`mtp.*`).
-## Рецепт
 ```python
 from transformers import AutoProcessor, Qwen3_5MoeForConditionalGeneration
@@ -65,15 +70,23 @@ model.save_pretrained("out")
 processor.save_pretrained("out")
 ```
-## Запуск в vLLM
-Спекулятивное декодирование (MTP) намеренно **не** используется.
-Мышление (`<think>`) выключается на уровне шаблона чата.
 ```bash
 vllm serve cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic \
-  --served-model-name joyfox-test \
   --max-model-len 16384 --kv-cache-dtype fp8 \
   --limit-mm-per-prompt '{"image": 0, "audio": 0, "video": 0}' \
   --default-chat-template-kwargs '{"enable_thinking": false}'
 ```

 library_name: transformers
 pipeline_tag: image-text-to-text
 base_model: joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive
+base_model_relation: quantized
 tags:
 - fp8
 - llm-compressor
 # JoyFox-Qwen3.6-35B-A3B-RP-Aggressive — FP8 Dynamic
+FP8_DYNAMIC (W8A8) quantization of
 [joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive](https://huggingface.co/joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive)
+(architecture `Qwen3_5MoeForConditionalGeneration`, base model `Qwen/Qwen3.6-35B-A3B`) —
+a 35B-A3B MoE roleplay finetune.
+* **Source:** `joyfox/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive`
+* **License:** apache-2.0 (inherited from the source and base models)
+* **Tool:** [llm-compressor](https://github.com/vllm-project/llm-compressor), data-free one-shot
+* **Scheme:** FP8 per-channel weights (static), FP8 per-token activations (dynamic)
+* **Weights size:** 36.66 GiB (down from ~70 GiB BF16) — fits a single 96 GB GPU with ample room for KV cache
+## Kept in BF16
+The following modules were excluded from quantization: `lm_head`, embeddings,
+the vision tower (`model.visual.*`), the Gated DeltaNet linear-attention layers
+(`linear_attn.*`), MoE routers (`mlp.gate`), the shared-expert gate
+(`shared_expert_gate`) and the MTP head (`mtp.*`). The ignore list matches the
+reference `RedHatAI/Qwen3.6-35B-A3B-FP8-dynamic` quantization.
+## Recipe
 ```python
 from transformers import AutoProcessor, Qwen3_5MoeForConditionalGeneration
 processor.save_pretrained("out")
 ```
+## Serving with vLLM
 ```bash
 vllm serve cloud19/JoyFox-Qwen3.6-35B-A3B-RP-Aggressive-FP8-Dynamic \
+  --served-model-name joyfox \
+  --enable-chunked-prefill \
   --max-model-len 16384 --kv-cache-dtype fp8 \
   --limit-mm-per-prompt '{"image": 0, "audio": 0, "video": 0}' \
   --default-chat-template-kwargs '{"enable_thinking": false}'
 ```
+Notes:
+* **Chunked prefill is required.** The hybrid attention stack (Gated DeltaNet
+  linear-attention layers) uses vLLM's mamba cache in mode `align`; passing
+  `--no-enable-chunked-prefill` makes the engine refuse to start.
+* The MTP head is shipped as `model_mtp.safetensors`, but speculative decoding
+  is intentionally **not** used here.
+* Thinking (`<think>`) is disabled at the chat-template level via
+  `enable_thinking: false`.