Instructions to use kitaniai/clover-1-150b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kitaniai/clover-1-150b with Transformers:

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

pipe = pipeline("image-text-to-text", model="kitaniai/clover-1-150b", trust_remote_code=True)
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("kitaniai/clover-1-150b", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use kitaniai/clover-1-150b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kitaniai/clover-1-150b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kitaniai/clover-1-150b",
		"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/kitaniai/clover-1-150b

SGLang

How to use kitaniai/clover-1-150b 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 "kitaniai/clover-1-150b" \
    --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": "kitaniai/clover-1-150b",
		"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 "kitaniai/clover-1-150b" \
        --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": "kitaniai/clover-1-150b",
		"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 kitaniai/clover-1-150b with Docker Model Runner:
```
docker model run hf.co/kitaniai/clover-1-150b
```

Clover 1 150B

Developer: Kitani
Architecture: clover_1_moe
Base model: Qwen/Qwen3.8-27B

Clover 1 150B is a Kitani derivative of Qwen3.8-27B. It preserves the trained Qwen3.8 tokenizer, multimodal vision tower, hybrid Gated DeltaNet/full-attention backbone, embeddings and language head, while replacing every language-model FFN with a custom sparse mixture-of-experts layer.

Clover changes

8 full-size routed experts per language layer
top-1 expert routing
fp32 learned router with per-expert bias
training-time router jitter
Switch-style load-balancing objective
router z-loss
optional entropy regularization
custom Hugging Face architecture: clover_1_moe

Every expert is initialized from its corresponding dense Qwen3.8 FFN. This is a function-preserving upcycling initialization; capability gains require continued training so the experts and router specialize.

Load

from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained(
    "Kitani/Clover-1-150B",
    trust_remote_code=True,
)
model = AutoModelForImageTextToText.from_pretrained(
    "Kitani/Clover-1-150B",
    trust_remote_code=True,
    dtype="auto",
    device_map="auto",
)

Attribution

Based on Qwen3.8-27B by the Qwen team. Apache-2.0 attribution and modification notices are retained. Clover branding does not imply that Qwen/Alibaba authored or endorsed Kitani's modifications.

Downloads last month: 5

Model tree for kitaniai/clover-1-150b

Base model

Qwen/Qwen3.8-27B

Finetuned

(755)

this model