Llama-v3.2-3B-Instruct: Optimized for Qualcomm Devices

Llama 3 is a family of LLMs. The model is quantized to w4a16 (4-bit weights and 16-bit activations) and part of the model is quantized to w8a16 (8-bit weights and 16-bit activations) making it suitable for on-device deployment.

This is based on the implementation of Llama-v3.2-3B-Instruct found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.

Deploying Llama-v3.2-3B-Instruct on-device

Follow the GenieX quickstart to install GenieX and deploy the model on a target device.

You'll need to export the model artifact using the steps below, then follow Run a Local Model with GenieX.

See the LLM-on-Genie tutorial to run with the Genie runtime. Note: Genie support will be deprecated soon.

Getting Started

Due to licensing restrictions, we cannot distribute pre-exported model assets for this model. Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:

Custom weights (e.g., fine-tuned checkpoints)
Custom input shapes
Target device and runtime configurations

See our repository for Llama-v3.2-3B-Instruct on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

Response Rate: Rate of response generation after the first response token.
Supported languages: English.
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).

Performance Summary

Model	Runtime	Precision	Chipset	Context Length	Response Rate (tokens per second)	Time To First Token (range, seconds)
Llama-v3.2-3B-Instruct	GENIE	w4	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4096	16.239038467407227	0.132916 - 4.253312
Llama-v3.2-3B-Instruct	GENIE	w4	Snapdragon® 8 Elite For Galaxy Mobile	4096	13.809293746948242	0.142042 - 4.545344
Llama-v3.2-3B-Instruct	GENIE	w4	Snapdragon® X Elite	4096	7.1577348709106445	0.25013599999999997 - 8.004351999999999
Llama-v3.2-3B-Instruct	GENIE	w4	Qualcomm® SA8295P	1024	3.523	0.37311700000000003 - 2.9849360000000003
Llama-v3.2-3B-Instruct	GENIE	w4	Qualcomm® Dragonwing™ IQ-X7181	4096	7.1577348709106445	0.25013599999999997 - 8.004351999999999
Llama-v3.2-3B-Instruct	GENIE	w4	Qualcomm® Dragonwing™ Q-8750	4096	13.809293746948242	0.142042 - 4.545344
Llama-v3.2-3B-Instruct	GENIE	w4a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4096	30.098724365234375	0.07220399999999999 - 2.3105279999999997
Llama-v3.2-3B-Instruct	GENIE	w4a16	Snapdragon® 8 Elite For Galaxy Mobile	4096	28.026905059814453	0.076697 - 2.454304
Llama-v3.2-3B-Instruct	GENIE	w4a16	Snapdragon® X2 Elite	4096	42.45383071899414	0.08531 - 2.72992
Llama-v3.2-3B-Instruct	GENIE	w4a16	Snapdragon® X Elite	4096	11.322591781616211	0.119393 - 3.820576
Llama-v3.2-3B-Instruct	GENIE	w4a16	Qualcomm® SA8775P	4096	17.47	0.109614 - 3.507648
Llama-v3.2-3B-Instruct	GENIE	w4a16	Qualcomm® SA8650P	4096	17.47	0.109614 - 3.507648
Llama-v3.2-3B-Instruct	GENIE	w4a16	Qualcomm® SA8255P	4096	17.47	0.109614 - 3.507648
Llama-v3.2-3B-Instruct	GENIE	w4a16	Qualcomm® Dragonwing™ IQ-9075	4096	18.14849090576172	0.10404000000000001 - 3.3292800000000002
Llama-v3.2-3B-Instruct	GENIE	w4a16	Qualcomm® Dragonwing™ IQ-X7181	4096	11.322591781616211	0.119393 - 3.820576
Llama-v3.2-3B-Instruct	GENIE	w4a16	Qualcomm® Dragonwing™ Q-8750	4096	28.026905059814453	0.076697 - 2.454304
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4096	11.029498	0.1835288064516129 - 5.872921806451613
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Snapdragon® 8 Elite For Galaxy Mobile	4096	10.591679	0.18471429032258063 - 5.91085729032258
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Snapdragon® X2 Elite	4096	16.409295	0.16508099999999998 - 5.282591999999999
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Snapdragon® X Elite	4096	8.794368	0.274154 - 8.772928
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Qualcomm® Dragonwing™ IQ-8275	4096	8.431866	0.344866 - 11.035712
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Qualcomm® Dragonwing™ IQ-9075	4096	8.445666	0.345638 - 11.060416
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Qualcomm® Dragonwing™ IQ-X7181	4096	8.794368	0.274154 - 8.772928
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4	Qualcomm® Dragonwing™ Q-8750	4096	10.591679	0.18471429032258063 - 5.91085729032258
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4096	17.797098	0.09555706451612901 - 3.0578260645161284
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Snapdragon® 8 Elite For Galaxy Mobile	4096	16.868169	0.10970051612903226 - 3.510416516129032
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Snapdragon® X2 Elite	4096	42.751892	0.061876 - 1.980032
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Snapdragon® X Elite	4096	19.824873	0.129435 - 4.14192
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Qualcomm® Dragonwing™ IQ-8275	4096	16.759964	0.12714 - 4.06848
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Qualcomm® Dragonwing™ IQ-9075	4096	18.385306	0.12773 - 4.08736
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Qualcomm® Dragonwing™ IQ-X7181	4096	19.824873	0.129435 - 4.14192
Llama-v3.2-3B-Instruct	GENIEX_QAIRT	w4a16	Qualcomm® Dragonwing™ Q-8750	4096	16.868169	0.10970051612903226 - 3.510416516129032

License

The license for the original implementation of Llama-v3.2-3B-Instruct can be found here.

References

Community

Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
For questions or feedback please reach out to us.

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

Downloads last month: -; Downloads are not tracked for this model. How to track