Instructions to use nvidia/Mistral-NeMo-Minitron-8B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/Mistral-NeMo-Minitron-8B-Instruct with Transformers:

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

pipe = pipeline("text-generation", model="nvidia/Mistral-NeMo-Minitron-8B-Instruct")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nvidia/Mistral-NeMo-Minitron-8B-Instruct")
model = AutoModelForCausalLM.from_pretrained("nvidia/Mistral-NeMo-Minitron-8B-Instruct")
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 nvidia/Mistral-NeMo-Minitron-8B-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nvidia/Mistral-NeMo-Minitron-8B-Instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nvidia/Mistral-NeMo-Minitron-8B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nvidia/Mistral-NeMo-Minitron-8B-Instruct

SGLang

How to use nvidia/Mistral-NeMo-Minitron-8B-Instruct 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 "nvidia/Mistral-NeMo-Minitron-8B-Instruct" \
    --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": "nvidia/Mistral-NeMo-Minitron-8B-Instruct",
		"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 "nvidia/Mistral-NeMo-Minitron-8B-Instruct" \
        --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": "nvidia/Mistral-NeMo-Minitron-8B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nvidia/Mistral-NeMo-Minitron-8B-Instruct with Docker Model Runner:
```
docker model run hf.co/nvidia/Mistral-NeMo-Minitron-8B-Instruct
```

Stop token is missing in tokenizer vocab

by armin-cpl - opened Oct 11, 2024

Discussion

armin-cpl

Oct 11, 2024

The following is included as an instruction in the section "Prompt Format" of the model card:
We recommend using <extra_id_1> as a stop token.

However, the tokenizer vocab does not include a token for <extra_id_1> and the string tokenizes to multiple tokens [1060, 37600, 3384, 1095, 1049, 1062]. This breaks the model usage with the prompt template for chat.

suhara

NVIDIA org Oct 15, 2024

Please use stop_strings as in the examples in the model card. For example,

outputs = model.generate(tokenized_chat, stop_strings=["<extra_id_1>"], tokenizer=tokenizer)

Tibbnak

Oct 16, 2024

The following is included as an instruction in the section "Prompt Format" of the model card:
We recommend using <extra_id_1> as a stop token.

However, the tokenizer vocab does not include a token for <extra_id_1> and the string tokenizes to multiple tokens [1060, 37600, 3384, 1095, 1049, 1062]. This breaks the model usage with the prompt template for chat.

nvidia dgaf about open standards and never have. P-agg using their own snowflake tokens and not even including them in the tokenizer defines fits right in line with how they do business.

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