Instructions to use aixsatoshi/calm2-7b-chat-7b-moe with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use aixsatoshi/calm2-7b-chat-7b-moe with Transformers:

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

pipe = pipeline("text-generation", model="aixsatoshi/calm2-7b-chat-7b-moe")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("aixsatoshi/calm2-7b-chat-7b-moe")
model = AutoModelForCausalLM.from_pretrained("aixsatoshi/calm2-7b-chat-7b-moe", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use aixsatoshi/calm2-7b-chat-7b-moe with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "aixsatoshi/calm2-7b-chat-7b-moe"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "aixsatoshi/calm2-7b-chat-7b-moe",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/aixsatoshi/calm2-7b-chat-7b-moe

SGLang

How to use aixsatoshi/calm2-7b-chat-7b-moe 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 "aixsatoshi/calm2-7b-chat-7b-moe" \
    --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": "aixsatoshi/calm2-7b-chat-7b-moe",
		"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 "aixsatoshi/calm2-7b-chat-7b-moe" \
        --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": "aixsatoshi/calm2-7b-chat-7b-moe",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use aixsatoshi/calm2-7b-chat-7b-moe with Docker Model Runner:
```
docker model run hf.co/aixsatoshi/calm2-7b-chat-7b-moe
```

Model Card for Model ID

This model represents an advanced implementation of a Mixture of Experts (MoE) approach, where cyberagent/calm2-7b serves as the foundational base model, and cyberagent/calm2-7b-chat is incorporated as an chat model. The model is designed to combine the general-purpose language processing capabilities of the calm2-7b with the specialized conversational abilities of the calm2-7b-chat.

Model Details

The model uses the following expert models for generating responses:

Source Model: cyberagent/calm2-7b-chat
- Positive Prompts: ["USER: ", "ASSISTANT: "]
  - This source model is utilized to provide responses in a chat-based context, taking both user and assistant inputs into account.
Source Model: cyberagent/calm2-7b
- Positive Prompts: [""]
  - This source model contributes to generating responses without specific chat context, serving as a general-purpose language model.

Model size: 11.3B
Context length: 32768
Language(s): Japanese, English

Model Sources [optional]

Repository: https://huggingface.co/cyberagent/calm2-7b
Repository: https://huggingface.co/cyberagent/calm2-7b-chat

Limitations and Considerations

While this MoE model integrates the strengths of cyberagent/calm2-7b-chat and cyberagent/calm2-7b, it's important to note that it is an experimental model and has not been fine-tuned post-composition. As such, users are advised to perform their own tuning and optimization to adapt the model to their specific use cases and requirements.

Downloads last month: 14

Safetensors

Model size

11B params

Tensor type

BF16