File size: 8,636 Bytes
abd5cde | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 | ---
library_name: pytorch
license: other
tags:
- llm
- generative_ai
- android
pipeline_tag: text-generation
---
# Granite-4.0-Micro: Optimized for Qualcomm Devices
Granite 4.0 is a family of open language models from IBM designed for enterprise AI workloads including code generation, summarization, and retrieval-augmented generation.
This is based on the implementation of Granite-4.0-Micro found [here](https://huggingface.co/ibm-granite/granite-4.0-h-micro).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/granite_4_0_micro) 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 Granite-4.0-Micro on-device
Follow the [GenieX quickstart](https://geniex.aihub.qualcomm.com/en/get-started/quickstart) to install GenieX and deploy the model on a target device.
## Getting Started
There are two ways to deploy this model on your device:
### Option 1: Download Pre-Exported Models
Below are pre-exported model assets ready for deployment.
| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| GENIEX_LLAMACPP | q4_0 | Universal | | [Download](https://huggingface.co/unsloth/granite-4.0-h-micro-GGUF/resolve/main/granite-4.0-h-micro-Q4_0.gguf)
For more device-specific assets and performance metrics, visit **[Granite-4.0-Micro on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/granite_4_0_micro)**.
### Option 2: Export with Custom Configurations
Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/granite_4_0_micro) 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
This option is ideal if you need to customize the model beyond the default configuration provided here.
See our repository for [Granite-4.0-Micro on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/granite_4_0_micro) for usage instructions.
## Model Details
**Model Type:** Model_use_case.text_generation
**Model Stats:**
- Model architecture: Mamba-2 Hybrid architecture combining State Space Model (SSM) layers with Attention layers for efficient long-context processing.
- 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.
- 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)
|---|---|---|---|---|---|---
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Gen 5 Mobile | 512 | 25.719503 | 0.8499905 - 3.399962
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Gen 5 Mobile | 512 | 26.283972 | 0.8927765000000001 - 3.5711060000000003
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Gen 5 Mobile | 512 | 16.840913 | 0.23313175 - 0.932527
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Gen 5 Mobile | 4096 | 24.522214 | 1.1895051875 - 38.064166
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Gen 5 Mobile | 4096 | 22.047256 | 1.4127757187499999 - 45.208822999999995
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Gen 5 Mobile | 4096 | 15.315768 | 0.3090850625 - 9.890722
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Mobile | 512 | 25.130048 | 0.9481944999999999 - 3.7927779999999998
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Mobile | 512 | 26.288118 | 0.95749725 - 3.829989
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Mobile | 512 | 15.760148 | 0.30860325 - 1.234413
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Mobile | 4096 | 22.639083 | 1.1536551875 - 36.916966
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Mobile | 4096 | 23.049816 | 1.22555709375 - 39.217827
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® 8 Elite Mobile | 4096 | 14.515548 | 0.36028896875 - 11.529247
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® X Elite | 512 | 17.843282 | 0.69928525 - 2.797141
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® X Elite | 512 | 13.447186 | 0.6960275 - 2.78411
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® X Elite | 512 | 11.275263 | 0.34302550000000004 - 1.3721020000000002
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® X Elite | 4096 | 25.288631 | 0.51570025 - 16.502408
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® X Elite | 4096 | 24.661293 | 0.54267315625 - 17.365541
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Snapdragon® X Elite | 4096 | 11.122384 | 0.37017084375000003 - 11.845467000000001
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ IQ-X7181 | 512 | 17.843282 | 0.69928525 - 2.797141
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ IQ-X7181 | 512 | 13.447186 | 0.6960275 - 2.78411
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ IQ-X7181 | 512 | 11.275263 | 0.34302550000000004 - 1.3721020000000002
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ Q-8750 | 512 | 25.130048 | 0.9481944999999999 - 3.7927779999999998
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ Q-8750 | 512 | 26.288118 | 0.95749725 - 3.829989
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ Q-8750 | 512 | 15.760148 | 0.30860325 - 1.234413
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ IQ-X7181 | 4096 | 25.288631 | 0.51570025 - 16.502408
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ IQ-X7181 | 4096 | 24.661293 | 0.54267315625 - 17.365541
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ IQ-X7181 | 4096 | 11.122384 | 0.37017084375000003 - 11.845467000000001
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ Q-8750 | 4096 | 22.639083 | 1.1536551875 - 36.916966
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ Q-8750 | 4096 | 23.049816 | 1.22555709375 - 39.217827
| Granite-4.0-Micro | GENIEX_LLAMACPP | q4_0 | Qualcomm® Dragonwing™ Q-8750 | 4096 | 14.515548 | 0.36028896875 - 11.529247
## License
* The license for the original implementation of Granite-4.0-Micro can be found
[here](https://github.com/ibm-granite/granite-4.0-language-models/blob/main/LICENSE).
## References
* [Granite 4.0](https://www.ibm.com/granite/docs/models/granite)
* [Source Model Implementation](https://huggingface.co/ibm-granite/granite-4.0-h-micro)
## 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
|