Instructions to use state-spaces/mamba-370m-hf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use state-spaces/mamba-370m-hf with Transformers:

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

pipe = pipeline("text-generation", model="state-spaces/mamba-370m-hf")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("state-spaces/mamba-370m-hf")
model = AutoModelForCausalLM.from_pretrained("state-spaces/mamba-370m-hf", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use state-spaces/mamba-370m-hf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "state-spaces/mamba-370m-hf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "state-spaces/mamba-370m-hf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/state-spaces/mamba-370m-hf

SGLang

How to use state-spaces/mamba-370m-hf 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 "state-spaces/mamba-370m-hf" \
    --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": "state-spaces/mamba-370m-hf",
		"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 "state-spaces/mamba-370m-hf" \
        --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": "state-spaces/mamba-370m-hf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use state-spaces/mamba-370m-hf with Docker Model Runner:
```
docker model run hf.co/state-spaces/mamba-370m-hf
```

Can `MambaForCausalLM` be used directly for training instead of `AutoModelForCausalLM`?

by LumenscopeAI - opened Apr 16, 2024

Discussion

LumenscopeAI

Apr 16, 2024

Hello,

I'm currently working with the transformers library to train a model on causal language modeling tasks using the MambaForCausalLM class. However, I've noticed that the typical approach to training in the library uses AutoModelForCausalLM to load the model for training, and I'm wondering if it's possible and recommended to use MambaForCausalLM directly for training instead.

Here is the code snippet I'm referring to:

from datasets import load_dataset
from trl import SFTTrainer
from peft import LoraConfig
from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments

# Model loading for training
tokenizer = AutoTokenizer.from_pretrained("state-spaces/mamba-370m-hf")
model = AutoModelForCausalLM.from_pretrained("state-spaces/mamba-370m-hf")

In inference, I successfully use MambaForCausalLM as follows:

from transformers import MambaConfig, MambaForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("/home/SLLaMA/mamba-370m-hf")
model = MambaForCausalLM.from_pretrained("/home/SLLaMA/mamba-370m-hf")

Could you clarify if using MambaForCausalLM for training is supported and if there are any additional configurations required for this?

Thank you for your assistance.

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