Instructions to use Inv/MoECPM-Untrained-4x2b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Inv/MoECPM-Untrained-4x2b with Transformers:

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

pipe = pipeline("text-generation", model="Inv/MoECPM-Untrained-4x2b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Inv/MoECPM-Untrained-4x2b")
model = AutoModelForCausalLM.from_pretrained("Inv/MoECPM-Untrained-4x2b")
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 Inv/MoECPM-Untrained-4x2b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Inv/MoECPM-Untrained-4x2b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Inv/MoECPM-Untrained-4x2b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Inv/MoECPM-Untrained-4x2b

SGLang

How to use Inv/MoECPM-Untrained-4x2b 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 "Inv/MoECPM-Untrained-4x2b" \
    --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": "Inv/MoECPM-Untrained-4x2b",
		"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 "Inv/MoECPM-Untrained-4x2b" \
        --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": "Inv/MoECPM-Untrained-4x2b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Inv/MoECPM-Untrained-4x2b with Docker Model Runner:
```
docker model run hf.co/Inv/MoECPM-Untrained-4x2b
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

MoECPM Untrained 4x2b

Model Details

Model Description

A MoE model out of 4 MiniCPM-2B-sft models. Intended to be trained. This version probably does not perform well (if it works at all, lol. I haven't tested it).

Uses

Training

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	53.51
AI2 Reasoning Challenge (25-Shot)	46.76
HellaSwag (10-Shot)	72.58
MMLU (5-Shot)	53.21
TruthfulQA (0-shot)	38.41
Winogrande (5-shot)	65.51
GSM8k (5-shot)	44.58

Downloads last month: 105

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for Inv/MoECPM-Untrained-4x2b

Base model

openbmb/MiniCPM-2B-sft-bf16-llama-format

Finetuned

(2)

this model

Quantizations

1 model

Evaluation results

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

46.760
normalized accuracy on HellaSwag (10-Shot)
validation set Open LLM Leaderboard

72.580
accuracy on MMLU (5-Shot)
test set Open LLM Leaderboard

53.210
mc2 on TruthfulQA (0-shot)
validation set Open LLM Leaderboard

38.410
accuracy on Winogrande (5-shot)
validation set Open LLM Leaderboard

65.510
accuracy on GSM8k (5-shot)
test set Open LLM Leaderboard

44.580