Instructions to use RedHatAI/Mistral-Nemo-Instruct-2407-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RedHatAI/Mistral-Nemo-Instruct-2407-FP8 with Transformers:

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

pipe = pipeline("text-generation", model="RedHatAI/Mistral-Nemo-Instruct-2407-FP8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("RedHatAI/Mistral-Nemo-Instruct-2407-FP8")
model = AutoModelForCausalLM.from_pretrained("RedHatAI/Mistral-Nemo-Instruct-2407-FP8")
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 RedHatAI/Mistral-Nemo-Instruct-2407-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RedHatAI/Mistral-Nemo-Instruct-2407-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/Mistral-Nemo-Instruct-2407-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/RedHatAI/Mistral-Nemo-Instruct-2407-FP8

SGLang

How to use RedHatAI/Mistral-Nemo-Instruct-2407-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/Mistral-Nemo-Instruct-2407-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/Mistral-Nemo-Instruct-2407-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/Mistral-Nemo-Instruct-2407-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/Mistral-Nemo-Instruct-2407-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use RedHatAI/Mistral-Nemo-Instruct-2407-FP8 with Docker Model Runner:
```
docker model run hf.co/RedHatAI/Mistral-Nemo-Instruct-2407-FP8
```

Lin-K76 commited on Jul 19, 2024

Commit

2d29ebe

verified ·

1 Parent(s): cd7369f

Update README.md

Browse files

Files changed (1) hide show

README.md +24 -24

README.md CHANGED Viewed

@@ -10,7 +10,7 @@ language:
 # Mistral-Nemo-Instruct-2407-FP8
 ## Model Overview
-- **Model Architecture:** Meta-Llama-3
   - **Input:** Text
   - **Output:** Text
 - **Model Optimizations:**
@@ -24,7 +24,7 @@ language:
 - **Model Developers:** Neural Magic
 Quantized version of [Mistral-Nemo-Instruct-2407](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407).
-It achieves an average score of 68.22 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 68.71.
 ### Model Optimizations
@@ -78,7 +78,7 @@ from transformers import AutoTokenizer
 from auto_fp8 import AutoFP8ForCausalLM, BaseQuantizeConfig
-pretrained_model_dir = "meta-llama/Mistral-Nemo-Instruct-2407"
 quantized_model_dir = "Mistral-Nemo-Instruct-2407-FP8"
 tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True, model_max_length=4096)
@@ -130,71 +130,71 @@ lm_eval \
   <tr>
    <td>MMLU (5-shot)
    </td>
-   <td>66.60
    </td>
-   <td>66.27
    </td>
-   <td>99.50%
    </td>
   </tr>
   <tr>
    <td>ARC Challenge (25-shot)
    </td>
-   <td>62.54
    </td>
-   <td>61.77
    </td>
-   <td>98.76%
    </td>
   </tr>
   <tr>
    <td>GSM-8K (5-shot, strict-match)
    </td>
-   <td>75.96
    </td>
-   <td>73.99
    </td>
-   <td>97.40%
    </td>
   </tr>
   <tr>
    <td>Hellaswag (10-shot)
    </td>
-   <td>78.83
    </td>
-   <td>78.56
    </td>
-   <td>99.65%
    </td>
   </tr>
   <tr>
    <td>Winogrande (5-shot)
    </td>
-   <td>75.93
    </td>
-   <td>76.40
    </td>
-   <td>100.6%
    </td>
   </tr>
   <tr>
    <td>TruthfulQA (0-shot)
    </td>
-   <td>52.44
    </td>
-   <td>52.35
    </td>
-   <td>99.82%
    </td>
   </tr>
   <tr>
    <td><strong>Average</strong>
    </td>
-   <td><strong>68.71</strong>
    </td>
-   <td><strong>68.22</strong>
    </td>
-   <td><strong>99.28%</strong>
    </td>
   </tr>
 </table>

 # Mistral-Nemo-Instruct-2407-FP8
 ## Model Overview
+- **Model Architecture:** Mistral-Nemo
   - **Input:** Text
   - **Output:** Text
 - **Model Optimizations:**
 - **Model Developers:** Neural Magic
 Quantized version of [Mistral-Nemo-Instruct-2407](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407).
+It achieves an average score of 71.28 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 71.61.
 ### Model Optimizations
 from auto_fp8 import AutoFP8ForCausalLM, BaseQuantizeConfig
+pretrained_model_dir = "mistralai/Mistral-Nemo-Instruct-2407"
 quantized_model_dir = "Mistral-Nemo-Instruct-2407-FP8"
 tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True, model_max_length=4096)
   <tr>
    <td>MMLU (5-shot)
    </td>
+   <td>68.35
    </td>
+   <td>68.50
    </td>
+   <td>100.2%
    </td>
   </tr>
   <tr>
    <td>ARC Challenge (25-shot)
    </td>
+   <td>65.53
    </td>
+   <td>64.68
    </td>
+   <td>98.70%
    </td>
   </tr>
   <tr>
    <td>GSM-8K (5-shot, strict-match)
    </td>
+   <td>74.45
    </td>
+   <td>73.01
    </td>
+   <td>98.06%
    </td>
   </tr>
   <tr>
    <td>Hellaswag (10-shot)
    </td>
+   <td>84.32
    </td>
+   <td>84.18
    </td>
+   <td>99.83%
    </td>
   </tr>
   <tr>
    <td>Winogrande (5-shot)
    </td>
+   <td>82.16
    </td>
+   <td>82.32
    </td>
+   <td>100.1%
    </td>
   </tr>
   <tr>
    <td>TruthfulQA (0-shot)
    </td>
+   <td>54.85
    </td>
+   <td>54.96
    </td>
+   <td>100.2%
    </td>
   </tr>
   <tr>
    <td><strong>Average</strong>
    </td>
+   <td><strong>71.61</strong>
    </td>
+   <td><strong>71.28</strong>
    </td>
+   <td><strong>99.53%</strong>
    </td>
   </tr>
 </table>