Instructions to use Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4 with Transformers:

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

pipe = pipeline("text-generation", model="Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4")
model = AutoModelForCausalLM.from_pretrained("Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4", device_map="auto")
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 Settings

vLLM

How to use Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4

SGLang

How to use Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4 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 "Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4" \
    --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": "Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4",
		"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 "Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4" \
        --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": "Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4 with Docker Model Runner:
```
docker model run hf.co/Kaleto/DeepSeek-R1-Distill-Llama-70B-NVFP4
```

Kaleto commited on May 20

Commit

d529b4d

verified ·

1 Parent(s): 21120da

Add Serving mode on Blackwell section — clarify weight-only FP4 reality on GB10+vLLM

Browse files

Files changed (1) hide show

README.md +6 -0

README.md CHANGED Viewed

@@ -31,6 +31,12 @@ To my knowledge this is the first publicly available NVFP4 of the DeepSeek-R1-Di
 ---
 ## Quick facts
 | | |

 ---
+## Serving mode on Blackwell (GB10)
+On DGX Spark / GB10 with vLLM, this model serves as **weight-only FP4**: the 4-bit NVFP4 weights are dequantized to BF16 for each matmul; activations stay BF16. vLLM 0.20.x has no FP4-activation GEMM kernel for Blackwell (sm_120/121), so the MARLIN-NVFP4 path is weight-only regardless of the `input_activations` field in `config.json` — verified by direct logit comparison (W4A4-config and W4A16-config produce bit-identical output on this stack). This is the standard, and currently highest-quality, NVFP4 serving mode on Spark. On an FP4-activation-capable stack (TensorRT-LLM, or a future vLLM with a Blackwell FP4 GEMM) the same weights could run as true W4A4.
+---
 ## Quick facts
 | | |