Instructions to use moonshotai/Kimi-K3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use moonshotai/Kimi-K3 with Transformers:

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

pipe = pipeline("image-text-to-text", model="moonshotai/Kimi-K3", 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)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("moonshotai/Kimi-K3", trust_remote_code=True, device_map="auto")

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use moonshotai/Kimi-K3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "moonshotai/Kimi-K3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "moonshotai/Kimi-K3",
		"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/moonshotai/Kimi-K3

SGLang

How to use moonshotai/Kimi-K3 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 "moonshotai/Kimi-K3" \
    --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": "moonshotai/Kimi-K3",
		"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 "moonshotai/Kimi-K3" \
        --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": "moonshotai/Kimi-K3",
		"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 moonshotai/Kimi-K3 with Docker Model Runner:
```
docker model run hf.co/moonshotai/Kimi-K3
```

Can i run this on my macbook?

#76

by hugginface536 - opened 1 day ago

Discussion

hugginface536

1 day ago

jackthejacks

1 day ago

dude u need a stack of maybe 8 h200 to run this...

ash-x-kingsley

1 day ago

dude u need a stack of maybe 8 h200 to run this...

It'd require 12x H200 or 9x B200.

SRZhu26

1 day ago

can i run this on a 2 gb ram raspberry pi? or maybe an apple watch? (gen 1) /s

OliverBright155

1 day ago

can i run this on a 2 gb ram raspberry pi? or maybe an apple watch? (gen 1) /s

I mean I got it running 12,400 t/s on a TI-84 graphing calculator (python ofc) so a raspberry pi should be easy game you got that

RKi714

1 day ago

Can you run it on a mac, yes. Check out his setup, he has a reddit post in the sub localLLM. google gavamedia/deltafin. He actually made a lot of process in 12 hours. this is just him running it on a mac 64gb of ram. Look forward to him and others speeding the process up.

hurler98

1 day ago

Link here https://github.com/gavamedia/deltafin . Seems possible since not 1.8T parameters are activated all the time, only like 30B I think. But man, the speed make it useless. It's at least one minute PER token. So like I guess a full answer might take a day or more...

RTHEMASTER

1 day ago

可以，当然，你需要一个mac集群，试试采用6台的m3u然后使用雷电5串联，我尝试过速度很慢并且上下文很小不可调整

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