Instructions to use Undi95/Mistral-11B-OmniMix9 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Undi95/Mistral-11B-OmniMix9 with Transformers:

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

pipe = pipeline("text-generation", model="Undi95/Mistral-11B-OmniMix9")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Undi95/Mistral-11B-OmniMix9")
model = AutoModelForCausalLM.from_pretrained("Undi95/Mistral-11B-OmniMix9")
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

vLLM

How to use Undi95/Mistral-11B-OmniMix9 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Undi95/Mistral-11B-OmniMix9"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Undi95/Mistral-11B-OmniMix9",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Undi95/Mistral-11B-OmniMix9

SGLang

How to use Undi95/Mistral-11B-OmniMix9 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 "Undi95/Mistral-11B-OmniMix9" \
    --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": "Undi95/Mistral-11B-OmniMix9",
		"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 "Undi95/Mistral-11B-OmniMix9" \
        --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": "Undi95/Mistral-11B-OmniMix9",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Undi95/Mistral-11B-OmniMix9 with Docker Model Runner:
```
docker model run hf.co/Undi95/Mistral-11B-OmniMix9
```

Don't mind those at the moment, I need to finetune them for RP, it's just some tests.

WARNING: This model specifically need EOS token I completely forgot to put on the json files, and need to check what was the right ones trough the mix. Please don't use it like this if you really want to review it.

slices:
  - sources:
    - model: "/content/drive/MyDrive/CC-v1.1-7B-bf16"
      layer_range: [0, 24]
  - sources:
    - model: "/content/drive/MyDrive/Zephyr-7B"
      layer_range: [8, 32]
merge_method: passthrough
dtype: bfloat16

================================================

slices:
  - sources:
      - model: "/content/drive/MyDrive/Mistral-11B-CC-Zephyr"
        layer_range: [0, 48]
      - model: Undi95/Mistral-11B-OpenOrcaPlatypus
        layer_range: [0, 48]
merge_method: slerp
base_model: "/content/drive/MyDrive/Mistral-11B-CC-Zephyr"
parameters:
  t:
    - value: 0.5 # fallback for rest of tensors
dtype: bfloat16

hf-causal-experimental (pretrained=/content/drive/MyDrive/Mistral-11B-Test), limit: None, provide_description: False, num_fewshot: 0, batch_size: 4

Task	Version	Metric	Value		Stderr
arc_challenge	0	acc	0.5623	±	0.0145
		acc_norm	0.5794	±	0.0144
arc_easy	0	acc	0.8354	±	0.0076
		acc_norm	0.8165	±	0.0079
hellaswag	0	acc	0.6389	±	0.0048
		acc_norm	0.8236	±	0.0038
piqa	0	acc	0.8139	±	0.0091
		acc_norm	0.8264	±	0.0088
truthfulqa_mc	1	mc1	0.3978	±	0.0171
		mc2	0.5607	±	0.0155
winogrande	0	acc	0.7451	±	0.0122

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	53.06
ARC (25-shot)	64.08
HellaSwag (10-shot)	84.24
MMLU (5-shot)	64.0
TruthfulQA (0-shot)	56.19
Winogrande (5-shot)	78.45
GSM8K (5-shot)	16.15
DROP (3-shot)	8.35

Downloads last month: 46

Safetensors

Model size

11B params

Tensor type

BF16