Instructions to use IlyasMoutawwakil/tiny-random-DeepseekV4-Flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use IlyasMoutawwakil/tiny-random-DeepseekV4-Flash with Transformers:

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

pipe = pipeline("text-generation", model="IlyasMoutawwakil/tiny-random-DeepseekV4-Flash")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("IlyasMoutawwakil/tiny-random-DeepseekV4-Flash")
model = AutoModelForCausalLM.from_pretrained("IlyasMoutawwakil/tiny-random-DeepseekV4-Flash", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use IlyasMoutawwakil/tiny-random-DeepseekV4-Flash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "IlyasMoutawwakil/tiny-random-DeepseekV4-Flash"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "IlyasMoutawwakil/tiny-random-DeepseekV4-Flash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/IlyasMoutawwakil/tiny-random-DeepseekV4-Flash

SGLang

How to use IlyasMoutawwakil/tiny-random-DeepseekV4-Flash 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 "IlyasMoutawwakil/tiny-random-DeepseekV4-Flash" \
    --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": "IlyasMoutawwakil/tiny-random-DeepseekV4-Flash",
		"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 "IlyasMoutawwakil/tiny-random-DeepseekV4-Flash" \
        --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": "IlyasMoutawwakil/tiny-random-DeepseekV4-Flash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use IlyasMoutawwakil/tiny-random-DeepseekV4-Flash with Docker Model Runner:
```
docker model run hf.co/IlyasMoutawwakil/tiny-random-DeepseekV4-Flash
```

tiny-random-DeepseekV4-Flash

A tiny random model for testing, shrunk from deepseek-ai/DeepSeek-V4-Flash: the same architecture, quantization config and checkpoint layout at test sizes. Its key patterns, dtypes and tensor ranks match the real checkpoint's (scripts/extract_layout.py).

DeepSeek's own key names: routed experts packed FP4 (I8 (out, in/2) + E8M0 .scale per 32, via the triton_kernels reference) over block-FP8 attention / shared experts / MTP projections (e4m3 + E8M0 .scale per 128x128), a hash-routed first layer, hyper-connections and an mtp.0 block.

reference/ holds the same weights dequantized to bf16, under the unquantized model's keys: the reference to compare logits against, so a test measures what the load path and kernels add, not the quantization itself.

The weights are random; the outputs mean nothing. scripts/ rebuilds it from the real checkpoint's config.json.

Downloads last month: 263

Safetensors

Model size

19.3M params

Tensor type

I64

F32

BF16

F8_E4M3