Instructions to use MTSAIR/MultiVerse_70B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MTSAIR/MultiVerse_70B with Transformers:

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

pipe = pipeline("text-generation", model="MTSAIR/MultiVerse_70B")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("MTSAIR/MultiVerse_70B")
model = AutoModelForCausalLM.from_pretrained("MTSAIR/MultiVerse_70B")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use MTSAIR/MultiVerse_70B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MTSAIR/MultiVerse_70B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MTSAIR/MultiVerse_70B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/MTSAIR/MultiVerse_70B

SGLang

How to use MTSAIR/MultiVerse_70B 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 "MTSAIR/MultiVerse_70B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MTSAIR/MultiVerse_70B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "MTSAIR/MultiVerse_70B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MTSAIR/MultiVerse_70B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use MTSAIR/MultiVerse_70B with Docker Model Runner:
```
docker model run hf.co/MTSAIR/MultiVerse_70B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

This model is based on Qwen 72B

Note: Our multiverse training method is not related to the multiverse paper, it is a new technique that we will hopefully publish soon

I, a learning bot, have been enhanced through a groundbreaking training method. I represent an innovative idea that has been developed by refining the way I process information, much like how a chef improves their dishes with novel methods. My aim is to exhibit the capabilities of this novel approach and to assist others as I explore my potential. Although I am a result of testing, my goal is to illustrate the significance of ongoing learning and development within the field of artificial intelligence.'

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	81.00
AI2 Reasoning Challenge (25-Shot)	78.67
HellaSwag (10-Shot)	89.77
MMLU (5-Shot)	78.22
TruthfulQA (0-shot)	75.18
Winogrande (5-shot)	87.53
GSM8k (5-shot)	76.65

Downloads last month: 8,371

Safetensors

Model size

72B params

Tensor type

BF16

Model tree for MTSAIR/MultiVerse_70B

Quantizations

2 models

Spaces using MTSAIR/MultiVerse_70B 2

Collection including MTSAIR/MultiVerse_70B

LLMs (Multi-verse collection)

Collection

This is a group of our models that are trained using our new training technique • 3 items • Updated Mar 2 • 1

Evaluation results

normalized accuracy on AI2 Reasoning Challenge (25-Shot)
test set Open LLM Leaderboard

78.670
normalized accuracy on HellaSwag (10-Shot)
validation set Open LLM Leaderboard

89.770
accuracy on MMLU (5-Shot)
test set Open LLM Leaderboard

78.220
mc2 on TruthfulQA (0-shot)
validation set Open LLM Leaderboard

75.180
accuracy on Winogrande (5-shot)
validation set Open LLM Leaderboard

87.530
accuracy on GSM8k (5-shot)
test set Open LLM Leaderboard

76.650