Instructions to use deepseek-ai/deepseek-moe-16b-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepseek-ai/deepseek-moe-16b-base with Transformers:

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

pipe = pipeline("text-generation", model="deepseek-ai/deepseek-moe-16b-base", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-moe-16b-base", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use deepseek-ai/deepseek-moe-16b-base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepseek-ai/deepseek-moe-16b-base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deepseek-ai/deepseek-moe-16b-base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/deepseek-ai/deepseek-moe-16b-base

SGLang

How to use deepseek-ai/deepseek-moe-16b-base 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 "deepseek-ai/deepseek-moe-16b-base" \
    --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": "deepseek-ai/deepseek-moe-16b-base",
		"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 "deepseek-ai/deepseek-moe-16b-base" \
        --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": "deepseek-ai/deepseek-moe-16b-base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use deepseek-ai/deepseek-moe-16b-base with Docker Model Runner:
```
docker model run hf.co/deepseek-ai/deepseek-moe-16b-base
```

Fix compatibility with transformers 5.0

by jbernloehr - opened Mar 14

base: refs/heads/main

←

from: refs/pr/8

Discussion Files changed

+10

-13

jbernloehr

Mar 14

Fix three breaking changes introduced by transformers>=5.0:

Remove is_torch_fx_available import — This function was removed from transformers.utils.import_utils in v5.0. The torch.fx wrapping of _prepare_4d_causal_attention_mask is no longer needed.
Fix _init_rope rope_scaling key access — transformers 5.0 renamed rope_scaling["type"] to rope_scaling["rope_type"] and may auto-populate rope_scaling even when the original config has it as null. Use .get() with fallback for both key names, default factor to 1.0, and fall back to default DeepseekRotaryEmbedding for unknown scaling types instead of raising.
Add **kwargs to forward() signatures — transformers 5.0 passes additional keyword arguments (e.g., loss_mask) to model forward methods. Without **kwargs, these cause TypeError.

Ref: https://github.com/huggingface/transformers/issues/44561

Fix compatibility with transformers 5.0788f8a88

Houso

Apr 2

I meet the same question. Hope them would fix it.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

Ready to merge

This branch is ready to get merged automatically.

· Sign up or log in to comment