Instructions to use SHSLab/Step-5-Preview-BF16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SHSLab/Step-5-Preview-BF16 with Transformers:

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

pipe = pipeline("image-text-to-text", model="SHSLab/Step-5-Preview-BF16", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("SHSLab/Step-5-Preview-BF16", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use SHSLab/Step-5-Preview-BF16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SHSLab/Step-5-Preview-BF16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SHSLab/Step-5-Preview-BF16",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/SHSLab/Step-5-Preview-BF16

SGLang

How to use SHSLab/Step-5-Preview-BF16 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 "SHSLab/Step-5-Preview-BF16" \
    --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": "SHSLab/Step-5-Preview-BF16",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "SHSLab/Step-5-Preview-BF16" \
        --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": "SHSLab/Step-5-Preview-BF16",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use SHSLab/Step-5-Preview-BF16 with Docker Model Runner:
```
docker model run hf.co/SHSLab/Step-5-Preview-BF16
```

Step-5-Preview-BF16 / config.json

Manusagents

Duplicate from TypeSafeAI/Step-5-Preview-BF16

91b76c4 4 days ago

Raw

History Blame Contribute Delete

12 kB

	{
	"architectures": [
	"MMGPTStepRoboticsForCausalLM"
	],
	"auto_map": {
	"AutoConfig": "configuration_step_robotics.StepRoboticsConfig"
	},
	"model_type": "step3p5v",
	"im_end_token": "<im_end>",
	"im_patch_token": "<im_patch>",
	"im_start_token": "<im_start>",
	"image_token_len": 169,
	"patch_token_len": 81,
	"image_token_id": 128001,
	"understand_projector_stride": 2,
	"use_im_start_end": "true",
	"vision_select_layer": -1,
	"projector_bias": false,
	"vision_config": {
	"model_type": "perception_encoder",
	"image_size": 728,
	"patch_size": 14,
	"width": 1536,
	"layers": 47,
	"heads": 16,
	"pool_type": "none",
	"output_dim": null,
	"use_cls_token": false,
	"ls_init_value": 0.1,
	"use_ln_post": false,
	"hidden_act": "quick_gelu"
	},
	"text_config": {
	"architectures": [
	"Step4ForCausalLM"
	],
	"model_type": "step4",
	"att_impl_type": "GQA",
	"torch_dtype": "bfloat16",
	"norm_dtype": "float32",
	"bos_token_id": 0,
	"eos_token_id": [
	1,
	2
	],
	"hidden_size": 4096,
	"intermediate_size": 13824,
	"num_hidden_layers": 92,
	"vocab_size": 128896,
	"rms_norm_eps": 1e-05,
	"fp32_residual_connection": true,
	"num_attention_heads": 64,
	"num_attention_groups": 4,
	"num_sliding_attention_heads": 64,
	"head_dim": 192,
	"head_dim_full_attention": 192,
	"head_dim_sliding_attention": 192,
	"num_attention_groups_full_attention": 4,
	"num_attention_groups_sliding_attention": 4,
	"rope_theta": [
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0,
	10000000.0,
	10000.0,
	10000.0,
	10000.0
	],
	"layer_types": [
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention",
	"full_attention",
	"sliding_attention",
	"sliding_attention",
	"sliding_attention"
	],
	"partial_rotary_factors": [
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0,
	0.3333333333333333,
	1.0,
	1.0,
	1.0
	],
	"sliding_window": 512,
	"attention_other_setting": {
	"attention_type": "sliding_attention",
	"num_attention_heads": 64,
	"num_attention_groups": 4,
	"head_dim": 192,
	"true_head_dim": 192
	},
	"use_head_wise_attn_gate": true,
	"num_nextn_predict_layers": 3,
	"use_moe": true,
	"moe_intermediate_size": 1536,
	"share_expert_dim": 1536,
	"moe_num_experts": 352,
	"moe_top_k": 8,
	"num_experts_per_tok": 8,
	"moe_layer_list": [
	3,
	4,
	5,
	6,
	7,
	8,
	9,
	10,
	11,
	12,
	13,
	14,
	15,
	16,
	17,
	18,
	19,
	20,
	21,
	22,
	23,
	24,
	25,
	26,
	27,
	28,
	29,
	30,
	31,
	32,
	33,
	34,
	35,
	36,
	37,
	38,
	39,
	40,
	41,
	42,
	43,
	44,
	45,
	46,
	47,
	48,
	49,
	50,
	51,
	52,
	53,
	54,
	55,
	56,
	57,
	58,
	59,
	60,
	61,
	62,
	63,
	64,
	65,
	66,
	67,
	68,
	69,
	70,
	71,
	72,
	73,
	74,
	75,
	76,
	77,
	78,
	79,
	80,
	81,
	82,
	83,
	84,
	85,
	86,
	87,
	88,
	89,
	90
	],
	"moe_layers_enum": "3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90",
	"use_moe_router_bias": true,
	"need_fp32_gate": true,
	"swiglu_limits": [
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0
	],
	"swiglu_limits_shared": [
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0,
	7.0
	],
	"max_seq_len": 1048576,
	"max_position_embeddings": 1048576,
	"norm_expert_weight": true,
	"moe_router_scaling_factor": 3.0,
	"rope_scaling": {
	"rope_type": "llama3",
	"factor": 1.0,
	"original_max_position_embeddings": 1048576,
	"low_freq_factor": 1.0,
	"high_freq_factor": 32.0
	},
	"yarn_only_types": [
	"full_attention"
	],
	"sparse_config": {
	"enabled": true,
	"proxy_dim": 256,
	"sparse_indexer_rope_dim": 32,
	"sparse_indexer_use_rope": true,
	"sparse_indexer_num_heads": 16,
	"sparse_indexer_num_k_heads": 1,
	"sparse_indexer_q_norm_type": "rmsnorm",
	"sparse_indexer_k_norm_type": "layernorm",
	"sparse_indexer_csa_z_norm_type": "none",
	"sparse_indexer_softmax_variant": "ssmax",
	"sparse_indexer_ssmax_s_granularity": "q_head",
	"num_provider_groups": 4,
	"topk": 512,
	"region_block_size": 8,
	"compression_method": "csa_block_compress",
	"attention_impl": "sparse_gqa",
	"apply_to_layer_types": [
	"full_attention"
	]
	}
	}
	}