Instructions to use MarsupialAI/Monstral-123B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MarsupialAI/Monstral-123B with Transformers:

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

pipe = pipeline("text-generation", model="MarsupialAI/Monstral-123B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("MarsupialAI/Monstral-123B")
model = AutoModelForCausalLM.from_pretrained("MarsupialAI/Monstral-123B", 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 MarsupialAI/Monstral-123B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MarsupialAI/Monstral-123B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MarsupialAI/Monstral-123B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/MarsupialAI/Monstral-123B

SGLang

How to use MarsupialAI/Monstral-123B 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 "MarsupialAI/Monstral-123B" \
    --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": "MarsupialAI/Monstral-123B",
		"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 "MarsupialAI/Monstral-123B" \
        --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": "MarsupialAI/Monstral-123B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use MarsupialAI/Monstral-123B with Docker Model Runner:
```
docker model run hf.co/MarsupialAI/Monstral-123B
```

Best model regarding personality but Major flaw

by Autumnlight - opened Oct 25, 2024

Discussion

Autumnlight

Oct 25, 2024

This is the best model by far when it comes to personality and character IQ.
Issue is that (tested 2q) it suffers from low memory IQ.
Seems like 8-11K is the sweet spot, it has really good backlog attention, follows the scenario very well etc. But after that it breaks totally apart, repeats actions, forgets its goal and becomes basically a character with dementia sadly.

TheDrummer

Oct 28, 2024

Behemoth can go far past 32K so...

ddh0

Oct 28, 2024

It's probably because you're using a Q2 quant

MarsupialAI

Owner Oct 30, 2024

As per our discord conversation, your problem is likely a combination of using a Q2 quant and 4bit KV. I've had no issues out to 32k, which is as far as I can go on my system with iQ3m and 16bit KV.

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