Instructions to use FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2 with Transformers:

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

pipe = pipeline("text-generation", model="FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2")
model = AutoModelForCausalLM.from_pretrained("FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2

SGLang

How to use FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2 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 "FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2" \
    --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": "FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2",
		"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 "FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2" \
        --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": "FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2 with Docker Model Runner:
```
docker model run hf.co/FastFlowLM/Deepseek-R1-Distill-Llama-8B-NPU2
```

FastFlowLM commited on Jul 31, 2025

Commit

8deccd0

verified ·

1 Parent(s): 1902964

Create README.md

Browse files

Files changed (1) hide show

README.md +71 -0

README.md ADDED Viewed

	@@ -0,0 +1,71 @@

+---
+license: mit
+language:
+- en
+library_name: transformers
+tags:
+- deepseek
+- distillation
+- llama
+- llama-compatible
+- deepseek-r1
+- text-generation
+- AMD
+- Ryzen
+- NPU
+pipeline_tag: text-generation
+base_model:
+- deepseek-ai/Deepseek-R1-Distill-Llama-8B
+---
+# 🔬 Deepseek R1 Distill LLaMA 8B – Optimized for FastFlowLM on AMD Ryzen™ AI NPU (XDNA2 Only)
+## Model Summary
+This model is based on **Deepseek R1 Distill LLaMA 8B**, a distilled version of a LLaMA-compatible architecture trained by Deepseek AI. It is designed for high-speed inference with low power consumption using the FastFlowLM runtime on Ryzen™ AI NPUs.
+> ✅ **Released under the MIT License**
+## 📝 License & Usage Terms
+### Base Model License
+- Released by Deepseek AI under the MIT License:
+  👉 https://huggingface.co/deepseek-ai/Deepseek-R1-Distill-Llama-8B
+- License permits:
+  - Commercial and non-commercial use
+  - Redistribution and modification
+  - No attribution requirement (though encouraged)
+### Redistribution Notice
+- This repository does **not** include original or modified Deepseek model weights.
+- You must download the base weights from Hugging Face directly:
+  👉 https://huggingface.co/deepseek-ai/Deepseek-R1-Distill-Llama-8B
+### If Fine-tuned
+If the model has been quantized or further fine-tuned:
+- **Base Model License**: MIT
+- **Derivative Weights License**: [e.g., MIT, CC-BY-NC-4.0, custom]
+- **Training Dataset License(s)**:
+  - [Dataset A] – [license]
+  - [Dataset B] – [license]
+Ensure compliance with the licenses of any datasets used in training.
+## Intended Use
+- **Best Suited For**: Local LLM inference, chat assistants, coding helpers, research
+- **Not Suited For**: Sensitive or safety-critical use cases without additional testing
+## Limitations & Risks
+- May exhibit factual inaccuracies or biases
+- Distillation may reduce generalization on edge cases
+- Performance depends on quantization strategy and runtime settings
+## Citation
+```bibtex
+@misc{deepseek2024r1,
+  title={DeepSeek R1: Distilled LLaMA-based Model},
+  author={Deepseek AI},
+  year={2024},
+  url={https://huggingface.co/deepseek-ai/Deepseek-R1-Distill-Llama-8B}
+```