Instructions to use stepfun-ai/Step-3.5-Flash-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use stepfun-ai/Step-3.5-Flash-FP8 with Transformers:

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

pipe = pipeline("text-generation", model="stepfun-ai/Step-3.5-Flash-FP8", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("stepfun-ai/Step-3.5-Flash-FP8", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use stepfun-ai/Step-3.5-Flash-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "stepfun-ai/Step-3.5-Flash-FP8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "stepfun-ai/Step-3.5-Flash-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/stepfun-ai/Step-3.5-Flash-FP8

SGLang

How to use stepfun-ai/Step-3.5-Flash-FP8 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 "stepfun-ai/Step-3.5-Flash-FP8" \
    --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": "stepfun-ai/Step-3.5-Flash-FP8",
		"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 "stepfun-ai/Step-3.5-Flash-FP8" \
        --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": "stepfun-ai/Step-3.5-Flash-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use stepfun-ai/Step-3.5-Flash-FP8 with Docker Model Runner:
```
docker model run hf.co/stepfun-ai/Step-3.5-Flash-FP8
```

AttributeError: 'Step3p5Config' object has no attribute 'tie_word_embeddings'

by 1337Code - opened Feb 11

Discussion

1337Code

Feb 11

Lastest VLLM Nightly build, model will no longer start/load.

version 0.16.0rc2.dev75+ga9ecdc01d

0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783] Traceback (most recent call last):
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 754, in worker_main
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     worker = WorkerProc(*args, **kwargs)
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 580, in __init__
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     self.worker.load_model()
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu_worker.py", line 294, in load_model
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     self.model_runner.load_model(eep_scale_up=eep_scale_up)
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu_model_runner.py", line 4143, in load_model
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     self.model = model_loader.load_model(
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]                  ^^^^^^^^^^^^^^^^^^^^^^^^
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/model_loader/base_loader.py", line 54, in load_model
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     model = initialize_model(
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]             ^^^^^^^^^^^^^^^^^
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/model_loader/utils.py", line 54, in initialize_model
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     model = model_class(vllm_config=vllm_config, prefix=prefix)
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/step3p5.py", line 772, in __init__
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     self.model = Step3p5Model(
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]                  ^^^^^^^^^^^^^
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/compilation/decorators.py", line 305, in __init__
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     old_init(self, **kwargs)
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/step3p5.py", line 570, in __init__
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     config.tie_word_embeddings and get_pp_group().is_last_rank
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     ^^^^^^^^^^^^^^^^^^^^^^^^^^
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]   File "/usr/local/lib/python3.12/dist-packages/transformers/configuration_utils.py", line 164, in __getattribute__
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]     return super().__getattribute__(key)
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
0|vllm-swap  | (Worker_PP2 pid=209) ERROR 02-11 21:03:11 [multiproc_executor.py:783] AttributeError: 'Step3p5Config' object has no attribute 'tie_word_embeddings'

Tingdan

StepFun org Feb 12

Please provide the deployment commands. Is PP parallelism enabled in the current deployment?

1337Code

Feb 12

Yes PP 3

vllm/vllm-openai:nightly \
  --model /models/step35flash \
  --served-model-name step3p5-flash \
  --host 0.0.0.0 \
  --port 5000 \
  --max-model-len 168000 \
  --pipeline-parallel-size 3 \
  --gpu-memory-utilization 0.96 \
  --max-num-seqs 2 \
  --tool-call-parser step3p5 \
  --reasoning-parser step3p5 \
  --enable-auto-tool-choice \
  --enable-prefix-caching \
  --optimization-level 3 \
  --trust-remote-code \
  --disable-cascade-attn \
  --quantization fp8

1337Code

Mar 7

Hello?

Tingdan

StepFun org Mar 9

Sorry about the missing parameters. The config has been updated; please refer to this PR for the changes.

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