Instructions to use kurogane/multiscreen_154M_tinystorys_vocab768 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kurogane/multiscreen_154M_tinystorys_vocab768 with Transformers:

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

pipe = pipeline("text-generation", model="kurogane/multiscreen_154M_tinystorys_vocab768", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("kurogane/multiscreen_154M_tinystorys_vocab768", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use kurogane/multiscreen_154M_tinystorys_vocab768 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kurogane/multiscreen_154M_tinystorys_vocab768"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kurogane/multiscreen_154M_tinystorys_vocab768",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/kurogane/multiscreen_154M_tinystorys_vocab768

SGLang

How to use kurogane/multiscreen_154M_tinystorys_vocab768 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 "kurogane/multiscreen_154M_tinystorys_vocab768" \
    --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": "kurogane/multiscreen_154M_tinystorys_vocab768",
		"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 "kurogane/multiscreen_154M_tinystorys_vocab768" \
        --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": "kurogane/multiscreen_154M_tinystorys_vocab768",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use kurogane/multiscreen_154M_tinystorys_vocab768 with Docker Model Runner:
```
docker model run hf.co/kurogane/multiscreen_154M_tinystorys_vocab768
```

multiscreen_154M_tinystorys_vocab768 / __init__.py

kurogane

Upload 5 files

61e43e6 verified 4 months ago

Raw

History Blame Contribute Delete

1.94 kB

	"""Transformers-compatible Multiscreen implementation.

	This package ports the core architecture from ``dieOD/multiscreen-pytorch`` to
	Hugging Face Transformers-style ``PreTrainedConfig`` / ``PreTrainedModel``
	classes.
	"""

	from .configuration_multiscreen import MultiscreenConfig
	from .compile_utils import find_msvc_cl, load_vcvars_env, setup_compile_env
	from .data import PackedTextDataset
	from .modeling_multiscreen import (
	GatedScreeningBlock,
	MultiscreenForCausalLM,
	MultiscreenLayer,
	MultiscreenModel,
	MultiscreenPreTrainedModel,
	ScreeningCache,
	convert_original_state_dict_for_causal_lm,
	convert_original_state_dict_for_model,
	)

	__version__ = "0.1.2"

	__all__ = [
	"MultiscreenConfig",
	"MultiscreenPreTrainedModel",
	"MultiscreenModel",
	"MultiscreenForCausalLM",
	"MultiscreenLayer",
	"GatedScreeningBlock",
	"ScreeningCache",
	"convert_original_state_dict_for_causal_lm",
	"convert_original_state_dict_for_model",
	"PackedTextDataset",
	"find_msvc_cl",
	"load_vcvars_env",
	"setup_compile_env",
	"register_multiscreen_auto_classes",
	]


	def register_multiscreen_auto_classes() -> None:
	"""Register Multiscreen with Transformers auto classes in this process.

	Use this when loading local checkpoints without ``trust_remote_code`` and
	without installing the model into a Transformers source tree::

	from multiscreen_transformers import register_multiscreen_auto_classes
	register_multiscreen_auto_classes()

	from transformers import AutoModelForCausalLM
	model = AutoModelForCausalLM.from_pretrained("./checkpoint")
	"""

	from transformers import AutoConfig, AutoModel, AutoModelForCausalLM

	AutoConfig.register(MultiscreenConfig.model_type, MultiscreenConfig)
	AutoModel.register(MultiscreenConfig, MultiscreenModel)
	AutoModelForCausalLM.register(MultiscreenConfig, MultiscreenForCausalLM)