Instructions to use RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8 with Transformers:

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

pipe = pipeline("text-generation", model="RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8")
model = AutoModelForCausalLM.from_pretrained("RedHatAI/Meta-Llama-3.1-405B-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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RedHatAI/Meta-Llama-3.1-405B-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": "RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8

SGLang

How to use RedHatAI/Meta-Llama-3.1-405B-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 "RedHatAI/Meta-Llama-3.1-405B-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": "RedHatAI/Meta-Llama-3.1-405B-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 "RedHatAI/Meta-Llama-3.1-405B-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": "RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8 with Docker Model Runner:
```
docker model run hf.co/RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8
```

Lin-K76 commited on Aug 22, 2024

Commit

ce3efea

verified ·

1 Parent(s): 8132006

Update README.md

Browse files

Files changed (1) hide show

README.md +19 -19

README.md CHANGED Viewed

@@ -33,7 +33,7 @@ base_model: meta-llama/Meta-Llama-3.1-405B-Instruct
 - **Model Developers:** Neural Magic
 Quantized version of [Meta-Llama-3.1-405B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct) with the updated 8 kv-heads.
-It achieves an average score of 86.60 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 86.79.
 ### Model Optimizations
@@ -118,11 +118,11 @@ model_stub = "meta-llama/Meta-Llama-3.1-405B-Instruct"
 model_name = model_stub.split("/")[-1]
 device_map = calculate_offload_device_map(
-    model_stub, reserve_for_hessians=False, num_gpus=8, torch_dtype=torch.float16
 )
 model = SparseAutoModelForCausalLM.from_pretrained(
-    model_stub, torch_dtype=torch.float16, device_map=device_map
 )
 tokenizer = AutoTokenizer.from_pretrained(model_stub)
@@ -193,9 +193,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>87.41
    </td>
-   <td>87.05
    </td>
-   <td>99.59%
    </td>
   </tr>
   <tr>
@@ -203,9 +203,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>88.11
    </td>
-   <td>87.87
    </td>
-   <td>99.73%
    </td>
   </tr>
   <tr>
@@ -213,9 +213,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>94.97
    </td>
-   <td>94.97
    </td>
-   <td>100.0%
    </td>
   </tr>
   <tr>
@@ -223,9 +223,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>95.98
    </td>
-   <td>95.83
    </td>
-   <td>99.84%
    </td>
   </tr>
   <tr>
@@ -233,9 +233,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>88.54
    </td>
-   <td>88.11
    </td>
-   <td>99.51%
    </td>
   </tr>
   <tr>
@@ -243,9 +243,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>87.21
    </td>
-   <td>87.77
    </td>
-   <td>100.6%
    </td>
   </tr>
   <tr>
@@ -253,9 +253,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>65.31
    </td>
-   <td>64.58
    </td>
-   <td>98.88%
    </td>
   </tr>
   <tr>
@@ -263,9 +263,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td><strong>86.79</strong>
    </td>
-   <td><strong>86.60</strong>
    </td>
-   <td><strong>99.74%</strong>
    </td>
   </tr>
 </table>

 - **Model Developers:** Neural Magic
 Quantized version of [Meta-Llama-3.1-405B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct) with the updated 8 kv-heads.
+It achieves an average score of 86.78 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 86.79.
 ### Model Optimizations
 model_name = model_stub.split("/")[-1]
 device_map = calculate_offload_device_map(
+    model_stub, reserve_for_hessians=False, num_gpus=8, torch_dtype="auto"
 )
 model = SparseAutoModelForCausalLM.from_pretrained(
+    model_stub, torch_dtype="auto", device_map=device_map
 )
 tokenizer = AutoTokenizer.from_pretrained(model_stub)
    </td>
    <td>87.41
    </td>
+   <td>87.41
    </td>
+   <td>100.0%
    </td>
   </tr>
   <tr>
    </td>
    <td>88.11
    </td>
+   <td>88.02
    </td>
+   <td>99.90%
    </td>
   </tr>
   <tr>
    </td>
    <td>94.97
    </td>
+   <td>94.88
    </td>
+   <td>99.91%
    </td>
   </tr>
   <tr>
    </td>
    <td>95.98
    </td>
+   <td>96.29
    </td>
+   <td>100.3%
    </td>
   </tr>
   <tr>
    </td>
    <td>88.54
    </td>
+   <td>88.54
    </td>
+   <td>100.0%
    </td>
   </tr>
   <tr>
    </td>
    <td>87.21
    </td>
+   <td>86.98
    </td>
+   <td>99.74%
    </td>
   </tr>
   <tr>
    </td>
    <td>65.31
    </td>
+   <td>65.33
    </td>
+   <td>100.0%
    </td>
   </tr>
   <tr>
    </td>
    <td><strong>86.79</strong>
    </td>
+   <td><strong>86.78</strong>
    </td>
+   <td><strong>99.99%</strong>
    </td>
   </tr>
 </table>