README.md

---
library_name: pytorch
license: other
tags:
- llm
- generative_ai
- android
pipeline_tag: text-generation

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/phi_3_5_mini_instruct/web-assets/model_demo.png)

# Phi-3.5-Mini-Instruct: Optimized for Qualcomm Devices

Phi-3.5-mini is a lightweight, state-of-the-art open model built upon datasets used for Phi-3 - synthetic data and filtered publicly available websites - with a focus on very high-quality, reasoning dense data. The model belongs to the Phi-3 model family and supports 128K token context length. The model underwent a rigorous enhancement process, incorporating both supervised fine-tuning, proximal policy optimization, and direct preference optimization to ensure precise instruction adherence and robust safety measures.

This is based on the implementation of Phi-3.5-Mini-Instruct found [here](https://huggingface.co/microsoft/Phi-3.5-mini-instruct).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/phi_3_5_mini_instruct) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Deploying Phi-3.5-mini-instruct on-device

Please follow the [LLM on-device deployment](https://github.com/quic/ai-hub-apps/tree/main/tutorials/llm_on_genie) tutorial.

## Getting Started
Download pre-exported model assets from **[Phi-3.5-Mini-Instruct on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/phi_3_5_mini_instruct)**.



## Model Details

**Model Type:** Model_use_case.text_generation

**Model Stats:**
- Input sequence length for Prompt Processor: 128
- Context length: 4096
- Number of parameters: 3.8B
- Precision: w4a16 + w8a16 (few layers)
- TTFT: Time To First Token is the time it takes to generate the first response token. This is expressed as a range because it varies based on the length of the prompt. The lower bound is for a short prompt (up to 128 tokens, i.e., one iteration of the prompt processor) and the upper bound is for a prompt using the full context length (4096 tokens).
- Response Rate: Rate of response generation after the first response token.

## Performance Summary
| Model | Runtime | Precision | Chipset | Context Length | Response Rate (tokens per second) | Time To First Token (range, seconds)
|---|---|---|---|---|---|---
| Phi-3.5-Mini-Instruct | QNN_CONTEXT_BINARY | w4a16 | Snapdragon® 8 Elite Mobile | 4096 | 14.73 | 0.1195948 - 3.8270336
| Phi-3.5-Mini-Instruct | QNN_CONTEXT_BINARY | w4a16 | Snapdragon® X Elite | 4096 | 6.2 | 0.185833 - 5.946656
| Phi-3.5-Mini-Instruct | QNN_CONTEXT_BINARY | w4a16 | Snapdragon® 8 Gen 3 Mobile | 4096 | 13.01 | 0.1469056 - 4.7009792

## License
* The license for the original implementation of Phi-3.5-Mini-Instruct can be found
  [here](https://huggingface.co/microsoft/Phi-3.5-mini-instruct/blob/main/LICENSE).

## References
* [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://arxiv.org/abs/2404.14219)
* [Source Model Implementation](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)

## 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

This 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