---
library_name: pytorch
license: apache-2.0
pipeline_tag: text-generation
tags:
- llm
- generative_ai
- quantized
- android

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mistral_7b_instruct_v0_3_quantized/web-assets/model_demo.png)

# Mistral-7B-Instruct-v0_3: Optimized for Mobile Deployment
## State-of-the-art large language model useful on a variety of language understanding and generation tasks

The Mistral-7B-Instruct-v0.3 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-7B-v0.3.

This model is an implementation of Mistral-7B-Instruct-v0_3 found [here]({source_repo}).
This repository provides scripts to run Mistral-7B-Instruct-v0_3 on Qualcomm® devices.
More details on model performance across various devices, can be found
[here](https://aihub.qualcomm.com/models/mistral_7b_instruct_v0_3_quantized).


### Model Details

- **Model Type:** Text generation
- **Model Stats:**
  - Number of parameters: 7.3B
  - Precision: w8a16
  - Num of key-value heads: 8
  - Information about the model: ['Prompt Processor and Token Generator are split into 4 parts each.', 'Each corresponding Prompt Processor and Token Generator share weights.']
  - Max context length: 4096
  - Prompt processor model size: 4.17 GB
  - Prompt processor input: 128 tokens + KVCache initialized with pad token
  - Prompt processor output: 128 output tokens + KVCache for token generator
  - Token generator model size: 4.17 GB
  - Token generator input: 1 input token + past KVCache
  - Token generator output: 1 output token + KVCache for next iteration
  - Decoding length: 4096
  - Use: Initiate conversation with prompt-processor and then token generator for subsequent iterations.

| Model | Device | Chipset | Target Runtime | Response Rate (tokens per second) | Time To First Token (range, seconds) | Tiny MMLU 
|---|---|---|---|---|---|---|
| Mistral-7B-Instruct-v0_3 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 10.73 | 0.18 - 5.79 | 58.85% | Use Export Script |

## Deploying Mistral 7B Instruct v3.0 on-device
  Please follow [this tutorial](https://github.com/quic/ai-hub-apps/tree/main/tutorials/llama)
  to compile QNN binaries and generate bundle assets to run [ChatApp on Windows](https://github.com/quic/ai-hub-apps/tree/main/apps/windows/cpp/ChatApp) and on Android powered by QNN-Genie.



## Installation

This model can be installed as a Python package via pip.

```bash
pip install qai-hub-models
```


## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.

With this API token, you can configure your client to run models on the cloud
hosted devices.
```bash
qai-hub configure --api_token API_TOKEN
```
Navigate to [docs](https://app.aihub.qualcomm.com/docs/) for more information.



## Demo on-device

The package contains a simple end-to-end demo that downloads pre-trained
weights and runs this model on a sample input.

```bash
python -m qai_hub_models.models.mistral_7b_instruct_v0_3_quantized.demo
```

The above demo runs a reference implementation of pre-processing, model
inference, and post processing.

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.mistral_7b_instruct_v0_3_quantized.demo
```


### Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
device. This script does the following:
* Performance check on-device on a cloud-hosted device
* Downloads compiled assets that can be deployed on-device for Android.
* Accuracy check between PyTorch and on-device outputs.

```bash
python -m qai_hub_models.models.mistral_7b_instruct_v0_3_quantized.export
```
```
Profiling Results
------------------------------------------------------------

Device                       : Snapdragon 8 Elite QRD (15)
Runtime                      : QNN                        
Response Rate (Tokens/Second): 10.73                      
Time to First Token (Seconds): (0.18, 5.79)               
```


## How does this work?

This [export script](https://aihub.qualcomm.com/models/mistral_7b_instruct_v0_3_quantized/qai_hub_models/models/Mistral-7B-Instruct-v0_3/export.py)
leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
on-device. Lets go through each step below in detail:

Step 1: **Upload compiled model**

Upload compiled models from `qai_hub_models.models.mistral_7b_instruct_v0_3_quantized` on hub.
```python
import torch

import qai_hub as hub
from qai_hub_models.models.mistral_7b_instruct_v0_3_quantized import Model

# Load the model
model = Model.from_precompiled()

model_promptprocessor_part1 = hub.upload_model(model.prompt_processor_part1.get_target_model_path())
model_promptprocessor_part2 = hub.upload_model(model.prompt_processor_part2.get_target_model_path())
model_promptprocessor_part3 = hub.upload_model(model.prompt_processor_part3.get_target_model_path())
model_promptprocessor_part4 = hub.upload_model(model.prompt_processor_part4.get_target_model_path())
model_tokengenerator_part1 = hub.upload_model(model.token_generator_part1.get_target_model_path())
model_tokengenerator_part2 = hub.upload_model(model.token_generator_part2.get_target_model_path())
model_tokengenerator_part3 = hub.upload_model(model.token_generator_part3.get_target_model_path())
model_tokengenerator_part4 = hub.upload_model(model.token_generator_part4.get_target_model_path())
```


Step 2: **Performance profiling on cloud-hosted device**

After uploading compiled models from step 1. Models can be profiled model on-device using the
`target_model`. Note that this scripts runs the model on a device automatically
provisioned in the cloud.  Once the job is submitted, you can navigate to a
provided job URL to view a variety of on-device performance metrics.
```python

# Device
device = hub.Device("Samsung Galaxy S23")
profile_job_promptprocessor_part1 = hub.submit_profile_job(
    model=model_promptprocessor_part1,
    device=device,
)
profile_job_promptprocessor_part2 = hub.submit_profile_job(
    model=model_promptprocessor_part2,
    device=device,
)
profile_job_promptprocessor_part3 = hub.submit_profile_job(
    model=model_promptprocessor_part3,
    device=device,
)
profile_job_promptprocessor_part4 = hub.submit_profile_job(
    model=model_promptprocessor_part4,
    device=device,
)
profile_job_tokengenerator_part1 = hub.submit_profile_job(
    model=model_tokengenerator_part1,
    device=device,
)
profile_job_tokengenerator_part2 = hub.submit_profile_job(
    model=model_tokengenerator_part2,
    device=device,
)
profile_job_tokengenerator_part3 = hub.submit_profile_job(
    model=model_tokengenerator_part3,
    device=device,
)
profile_job_tokengenerator_part4 = hub.submit_profile_job(
    model=model_tokengenerator_part4,
    device=device,
)

```

Step 3: **Verify on-device accuracy**

To verify the accuracy of the model on-device, you can run on-device inference
on sample input data on the same cloud hosted device.
```python

input_data_promptprocessor_part1 = model.prompt_processor_part1.sample_inputs()
inference_job_promptprocessor_part1 = hub.submit_inference_job(
    model=model_promptprocessor_part1,
    device=device,
    inputs=input_data_promptprocessor_part1,
)
on_device_output_promptprocessor_part1 = inference_job_promptprocessor_part1.download_output_data()

input_data_promptprocessor_part2 = model.prompt_processor_part2.sample_inputs()
inference_job_promptprocessor_part2 = hub.submit_inference_job(
    model=model_promptprocessor_part2,
    device=device,
    inputs=input_data_promptprocessor_part2,
)
on_device_output_promptprocessor_part2 = inference_job_promptprocessor_part2.download_output_data()

input_data_promptprocessor_part3 = model.prompt_processor_part3.sample_inputs()
inference_job_promptprocessor_part3 = hub.submit_inference_job(
    model=model_promptprocessor_part3,
    device=device,
    inputs=input_data_promptprocessor_part3,
)
on_device_output_promptprocessor_part3 = inference_job_promptprocessor_part3.download_output_data()

input_data_promptprocessor_part4 = model.prompt_processor_part4.sample_inputs()
inference_job_promptprocessor_part4 = hub.submit_inference_job(
    model=model_promptprocessor_part4,
    device=device,
    inputs=input_data_promptprocessor_part4,
)
on_device_output_promptprocessor_part4 = inference_job_promptprocessor_part4.download_output_data()

input_data_tokengenerator_part1 = model.token_generator_part1.sample_inputs()
inference_job_tokengenerator_part1 = hub.submit_inference_job(
    model=model_tokengenerator_part1,
    device=device,
    inputs=input_data_tokengenerator_part1,
)
on_device_output_tokengenerator_part1 = inference_job_tokengenerator_part1.download_output_data()

input_data_tokengenerator_part2 = model.token_generator_part2.sample_inputs()
inference_job_tokengenerator_part2 = hub.submit_inference_job(
    model=model_tokengenerator_part2,
    device=device,
    inputs=input_data_tokengenerator_part2,
)
on_device_output_tokengenerator_part2 = inference_job_tokengenerator_part2.download_output_data()

input_data_tokengenerator_part3 = model.token_generator_part3.sample_inputs()
inference_job_tokengenerator_part3 = hub.submit_inference_job(
    model=model_tokengenerator_part3,
    device=device,
    inputs=input_data_tokengenerator_part3,
)
on_device_output_tokengenerator_part3 = inference_job_tokengenerator_part3.download_output_data()

input_data_tokengenerator_part4 = model.token_generator_part4.sample_inputs()
inference_job_tokengenerator_part4 = hub.submit_inference_job(
    model=model_tokengenerator_part4,
    device=device,
    inputs=input_data_tokengenerator_part4,
)
on_device_output_tokengenerator_part4 = inference_job_tokengenerator_part4.download_output_data()

```
With the output of the model, you can compute like PSNR, relative errors or
spot check the output with expected output.

**Note**: This on-device profiling and inference requires access to Qualcomm®
AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).




## Deploying compiled model to Android


The models can be deployed using multiple runtimes:
- TensorFlow Lite (`.tflite` export): [This
  tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
  guide to deploy the .tflite model in an Android application.


- QNN ( `.so` / `.bin` export ): This [sample
  app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
provides instructions on how to use the `.so` shared library or `.bin` context binary in an Android application.


## View on Qualcomm® AI Hub
Get more details on Mistral-7B-Instruct-v0_3's performance across various devices [here](https://aihub.qualcomm.com/models/mistral_7b_instruct_v0_3_quantized).
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)


## License
* The license for the original implementation of Mistral-7B-Instruct-v0_3 can be found [here](https://github.com/mistralai/mistral-inference/blob/main/LICENSE).
* The license for the compiled assets for on-device deployment can be found [here](https://github.com/mistralai/mistral-inference/blob/main/LICENSE)



## References
* [Mistral 7B](https://arxiv.org/abs/2310.06825)
* [Source Model Implementation](https://github.com/mistralai/mistral-inference)



## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).


## Usage and Limitations

Model may not be used for or in connection with any of the following applications:

- Accessing essential private and public services and benefits;
- Administration of justice and democratic processes;
- Assessing or recognizing the emotional state of a person;
- Biometric and biometrics-based systems, including categorization of persons based on sensitive characteristics;
- Education and vocational training;
- Employment and workers management;
- Exploitation of the vulnerabilities of persons resulting in harmful behavior;
- General purpose social scoring;
- Law enforcement;
- Management and operation of critical infrastructure;
- Migration, asylum and border control management;
- Predictive policing;
- Real-time remote biometric identification in public spaces;
- Recommender systems of social media platforms;
- Scraping of facial images (from the internet or otherwise); and/or
- Subliminal manipulation