Instructions to use SparseLLM/DECO-0.1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SparseLLM/DECO-0.1B with Transformers:

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

pipe = pipeline("text-generation", model="SparseLLM/DECO-0.1B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("SparseLLM/DECO-0.1B", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use SparseLLM/DECO-0.1B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SparseLLM/DECO-0.1B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SparseLLM/DECO-0.1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/SparseLLM/DECO-0.1B

SGLang

How to use SparseLLM/DECO-0.1B 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 "SparseLLM/DECO-0.1B" \
    --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": "SparseLLM/DECO-0.1B",
		"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 "SparseLLM/DECO-0.1B" \
        --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": "SparseLLM/DECO-0.1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use SparseLLM/DECO-0.1B with Docker Model Runner:
```
docker model run hf.co/SparseLLM/DECO-0.1B
```

DECO-0.1B

File size: 1,776 Bytes

8b38b65

{
    "architectures": [
        "BlockFFNForCausalLM"
    ],
    "auto_map": {
        "AutoConfig": "configuration_blockffn.BlockFFNConfig",
        "AutoModel": "modeling_blockffn.BlockFFNModel",
        "AutoModelForCausalLM": "modeling_blockffn.BlockFFNForCausalLM"
    },
    "bos_token_id": 1,
    "eos_token_id": [
        2,
        73440
    ],
    "pad_token_id": 2,
    "hidden_act": "silu",
    "hidden_size": 768,
    "initializer_range": 0.1,
    "intermediate_size": 10240,
    "head_dim": 128,
    "max_position_embeddings": 4096,
    "model_type": "blockffn",
    "num_attention_heads": 6,
    "num_hidden_layers": 32,
    "num_key_value_heads": 2,
    "rms_norm_eps": 1e-05,
    "rope_scaling": null,
    "rope_theta": 10000.0,
    "torch_dtype": "bfloat16",
    "transformers_version": "4.36.0",
    "use_cache": true,
    "vocab_size": 73448,
    "use_mup": false,
    "num_experts": 42,
    "moe_ffn_hidden_size": 64,
    "moe_shared_expert_intermediate_size": 128,
    "moe_layer_freq": [
        0,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1
    ],
    "moe_router_dtype": "fp32",
    "router_act_func": "relu",
    "router_norm_type": "simple",
    "expert_act_func": "norm_silu",
    "expert_act_norm_type": "normal",
    "num_layers": 16,
    "ffn_hidden_size": 1920,
    "num_query_groups": 6,
    "norm_epsilon": 1e-05,
    "use_blockffn": true,
    "router_type": "topk",
    "moe_router_enable_expert_bias": false,
    "expert_not_gated": true,
    "moe_router_pre_softmax": false,
    "moe_router_topk": 2,
    "moe_router_topp": 0.5,
    "moe_router_score_function": "softmax",
    "moe_router_topk_scaling_factor": null
}