Instructions to use amd/Instella-MoE-16B-A3B-Think with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use amd/Instella-MoE-16B-A3B-Think with Transformers:

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

pipe = pipeline("text-generation", model="amd/Instella-MoE-16B-A3B-Think", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("amd/Instella-MoE-16B-A3B-Think", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("amd/Instella-MoE-16B-A3B-Think", trust_remote_code=True, device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use amd/Instella-MoE-16B-A3B-Think with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "amd/Instella-MoE-16B-A3B-Think"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "amd/Instella-MoE-16B-A3B-Think",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/amd/Instella-MoE-16B-A3B-Think

SGLang

How to use amd/Instella-MoE-16B-A3B-Think 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 "amd/Instella-MoE-16B-A3B-Think" \
    --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": "amd/Instella-MoE-16B-A3B-Think",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "amd/Instella-MoE-16B-A3B-Think" \
        --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": "amd/Instella-MoE-16B-A3B-Think",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use amd/Instella-MoE-16B-A3B-Think with Docker Model Runner:
```
docker model run hf.co/amd/Instella-MoE-16B-A3B-Think
```

Congratulations and thank you to AMD on the release!

by vmlinux - opened Jul 25

Discussion

vmlinux

Jul 25

•

edited Jul 25

Thanks for the open source release AMD.

I'm going to pull everything down related to this and do some hobby-learning this weekend!

TawusGames

Jul 29

•

edited Jul 30

It is not open source or openweights look license. Research only license.

edit: Nvidia release their data fully open source. You can create same model from stracth. Sooo FUCK AMD. I bought fully amd rig for ai purpose and i regret it. Rocm still problematic. Vulkan works thanks to open source community. Trellis 2.0 (cumesh) still not working on Amd. No amd llm for local inference. Then they are release a model with researchrail lisance and no dataset WTF!!!

Sdoh

Aug 3

It is not open source or openweights look license. Research only license.

edit: Nvidia release their data fully open source. You can create same model from stracth. Sooo FUCK AMD. I bought fully amd rig for ai purpose and i regret it. Rocm still problematic. Vulkan works thanks to open source community. Trellis 2.0 (cumesh) still not working on Amd. No amd llm for local inference. Then they are release a model with researchrail lisance and no dataset WTF!!!

Бро, надо быть благодарным AMD за их существование.
Надо быть благодарным всем за их труд и вклад.

Но компании слишком большие и инертные. Если ты не попадаешь в их круг цели - идешь к чёрту ты. И в большинстве случаев идёшь к чёрту именно ты. Реальность. Добро пожаловать.

Спасибо тебе за комментарий.
Спасибо AMD и Nvidia за то, что они посылают нас к чёртку. М, очень вкусно, спасибо.
Кушаем всё это с надеждой на лучшее.

TawusGames

Aug 8

It is not open source or openweights look license. Research only license.

edit: Nvidia release their data fully open source. You can create same model from stracth. Sooo FUCK AMD. I bought fully amd rig for ai purpose and i regret it. Rocm still problematic. Vulkan works thanks to open source community. Trellis 2.0 (cumesh) still not working on Amd. No amd llm for local inference. Then they are release a model with researchrail lisance and no dataset WTF!!!

Бро, надо быть благодарным AMD за их существование.
Надо быть благодарным всем за их труд и вклад.

Но компании слишком большие и инертные. Если ты не попадаешь в их круг цели - идешь к чёрту ты. И в большинстве случаев идёшь к чёрту именно ты. Реальность. Добро пожаловать.

Спасибо тебе за комментарий.
Спасибо AMD и Nvidia за то, что они посылают нас к чёртку. М, очень вкусно, спасибо.
Кушаем всё это с надеждой на лучшее.

Customer is always right!!!

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