Instructions to use nvidia/Llama-3.1-8B-Instruct-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/Llama-3.1-8B-Instruct-FP8 with Transformers:

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

pipe = pipeline("text-generation", model="nvidia/Llama-3.1-8B-Instruct-FP8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nvidia/Llama-3.1-8B-Instruct-FP8")
model = AutoModelForCausalLM.from_pretrained("nvidia/Llama-3.1-8B-Instruct-FP8", 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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nvidia/Llama-3.1-8B-Instruct-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nvidia/Llama-3.1-8B-Instruct-FP8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nvidia/Llama-3.1-8B-Instruct-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nvidia/Llama-3.1-8B-Instruct-FP8

SGLang

How to use nvidia/Llama-3.1-8B-Instruct-FP8 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 "nvidia/Llama-3.1-8B-Instruct-FP8" \
    --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": "nvidia/Llama-3.1-8B-Instruct-FP8",
		"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 "nvidia/Llama-3.1-8B-Instruct-FP8" \
        --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": "nvidia/Llama-3.1-8B-Instruct-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nvidia/Llama-3.1-8B-Instruct-FP8 with Docker Model Runner:
```
docker model run hf.co/nvidia/Llama-3.1-8B-Instruct-FP8
```

Update README.md

by omrialmog - opened Feb 26, 2025

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

+25

-18

Files changed (1) hide show

README.md +25 -18

README.md CHANGED Viewed

@@ -1,6 +1,9 @@
 ---
 base_model:
 - meta-llama/Llama-3.1-8B-Instruct
 ---
 # Model Overview
@@ -75,39 +78,37 @@ python examples/llama/convert_checkpoint.py --model_dir Llama-3.1-8B-Instruct-FP
 trtllm-build --checkpoint_dir /ckpt --output_dir /engine
 ```
-* Accuracy evaluation:
-1) Prepare the MMLU dataset:
-```sh
-mkdir data; wget https://people.eecs.berkeley.edu/~hendrycks/data.tar -O data/mmlu.tar
-tar -xf data/mmlu.tar -C data && mv data/data data/mmlu
-```
-2) Measure MMLU:
-```sh
-python examples/mmlu.py --engine_dir ./engine --tokenizer_dir Llama-3.1-8B-Instruct-FP8/ --test_trt_llm --data_dir data/mmlu
-```
 * Throughputs evaluation:
 Please refer to the [TensorRT-LLM benchmarking documentation](https://github.com/NVIDIA/TensorRT-LLM/blob/main/benchmarks/Suite.md) for details.
 ## Evaluation
-The accuracy (MMLU, 5-shot) and throughputs (tokens per second, TPS) benchmark results are presented in the table below:
 <table>
   <tr>
    <td><strong>Precision</strong>
    </td>
    <td><strong>MMLU</strong>
    </td>
    <td><strong>TPS</strong>
    </td>
   </tr>
   <tr>
-   <td>FP16
    </td>
-   <td>68.6
    </td>
    <td>8,579.93
    </td>
@@ -115,7 +116,13 @@ The accuracy (MMLU, 5-shot) and throughputs (tokens per second, TPS) benchmark r
   <tr>
    <td>FP8
    </td>
-   <td>68.3
    </td>
    <td>11,062.90
    </td>

 ---
 base_model:
 - meta-llama/Llama-3.1-8B-Instruct
+license: llama3.1
+pipeline_tag: text-generation
+library_name: transformers
 ---
 # Model Overview
 trtllm-build --checkpoint_dir /ckpt --output_dir /engine
 ```
 * Throughputs evaluation:
 Please refer to the [TensorRT-LLM benchmarking documentation](https://github.com/NVIDIA/TensorRT-LLM/blob/main/benchmarks/Suite.md) for details.
 ## Evaluation
 <table>
   <tr>
    <td><strong>Precision</strong>
    </td>
    <td><strong>MMLU</strong>
    </td>
+   <td><strong>GSM8K (CoT) </strong>
+   </td>
+   <td><strong>ARC Challenge</strong>
+   </td>
+   <td><strong>IFEVAL</strong>
+   </td>
    <td><strong>TPS</strong>
    </td>
   </tr>
   <tr>
+   <td>BF16
+   </td>
+   <td>69.4
+   </td>
+   <td>84.5
    </td>
+   <td>83.4
+   </td>
+   <td>80.4
    </td>
    <td>8,579.93
    </td>
   <tr>
    <td>FP8
    </td>
+   <td>68.7
+   </td>
+   <td>83.1
+   </td>
+   <td>83.3
+   </td>
+   <td>81.8
    </td>
    <td>11,062.90
    </td>