Instructions to use z-lab/gemma-4-31B-it-DFlash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use z-lab/gemma-4-31B-it-DFlash with Transformers:

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

pipe = pipeline("text-generation", model="z-lab/gemma-4-31B-it-DFlash")

# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("z-lab/gemma-4-31B-it-DFlash")
model = AutoModel.from_pretrained("z-lab/gemma-4-31B-it-DFlash")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use z-lab/gemma-4-31B-it-DFlash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "z-lab/gemma-4-31B-it-DFlash"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "z-lab/gemma-4-31B-it-DFlash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/z-lab/gemma-4-31B-it-DFlash

SGLang

How to use z-lab/gemma-4-31B-it-DFlash 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 "z-lab/gemma-4-31B-it-DFlash" \
    --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": "z-lab/gemma-4-31B-it-DFlash",
		"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 "z-lab/gemma-4-31B-it-DFlash" \
        --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": "z-lab/gemma-4-31B-it-DFlash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use z-lab/gemma-4-31B-it-DFlash with Docker Model Runner:
```
docker model run hf.co/z-lab/gemma-4-31B-it-DFlash
```

gemma-4-31B-it-DFlash

File size: 1,096 Bytes

f3b070b

{
  "architectures": [
    "DFlashDraftModel"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "block_size": 16,
  "bos_token_id": 2,
  "dflash_config": {
    "mask_token_id": 4,
    "target_layer_ids": [
      1,
      12,
      23,
      35,
      46,
      57
    ]
  },
  "dtype": "bfloat16",
  "eos_token_id": 1,
  "final_logit_softcapping": 30.0,
  "head_dim": 128,
  "hidden_act": "silu",
  "hidden_size": 5376,
  "initializer_range": 0.02,
  "intermediate_size": 10752,
  "layer_types": [
    "sliding_attention",
    "sliding_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention"
  ],
  "max_position_embeddings": 262144,
  "max_window_layers": 5,
  "model_type": "qwen3",
  "num_attention_heads": 64,
  "num_hidden_layers": 5,
  "num_key_value_heads": 8,
  "num_target_layers": 60,
  "pad_token_id": 0,
  "rms_norm_eps": 1e-06,
  "sliding_window": 2048,
  "tie_word_embeddings": true,
  "transformers_version": "5.6.0",
  "use_cache": true,
  "use_sliding_window": true,
  "vocab_size": 262144,
  "rope_theta": 1000000,
  "rope_scaling": null
}