README.md

---
library_name: pytorch
license: llama3
pipeline_tag: text-generation
tags:
- llm
- generative_ai
- quantized
- android

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/llama_v3_8b_chat_quantized/web-assets/model_demo.png)

# Llama-v3-8B-Chat: Optimized for Mobile Deployment
## State-of-the-art large language model useful on a variety of language understanding and generation tasks


Llama 3 is a family of LLMs. The "Chat" at the end indicates that the model is optimized for chatbot-like dialogue. 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. For Prompt and output length specified below, the time to first token is Llama-PromptProcessor-Quantized's latency and average time per addition token is Llama-TokenGenerator-Quantized's latency.

This model is an implementation of Posenet-Mobilenet found [here](https://github.com/meta-llama/llama3/tree/main). 


More details on model performance accross various devices, can be found [here](https://aihub.qualcomm.com/models/llama_v3_8b_chat_quantized).

### Model Details

- **Model Type:** Text generation
- **Model Stats:**
  - Input sequence length for Prompt Processor: 128
  - Context length: 4096
  - Number of parameters: 8B
  - Model size: 4.8GB
  - Precision: w4a16 + w8a16 (few layers)
  - Num of key-value heads: 8
  - Model-1 (Prompt Processor): Llama-PromptProcessor-Quantized
  - Prompt processor input: 128 tokens + position embeddings + attention mask + KV cache inputs
  - Prompt processor output: 128 output tokens + KV cache outputs
  - Model-2 (Token Generator): Llama-TokenGenerator-Quantized
  - Token generator input: 1 input token + position embeddings + attention mask + KV cache inputs
  - Token generator output: 1 output token + KV cache outputs
  - Use: Initiate conversation with prompt-processor and then token generator for subsequent iterations.
  - Minimum QNN SDK version required: 2.27.7
  - 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).
  - Response Rate: Rate of response generation after the first response token.

| Model | Device | Chipset | Target Runtime | Response Rate (tokens per second) | Time To First Token (range, seconds)
|---|---|---|---|---|---|
| Llama-v3-8B-Chat | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 12.9262 | 0.159383 - 5.100256 | -- | -- |
| Llama-v3-8B-Chat | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 10.0367 | 0.211644 - 6.772608 | -- | -- |

## Deploying Llama 3 on-device

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



## License
* The license for the original implementation of Llama-v3-8B-Chat can be found [here](https://github.com/facebookresearch/llama/blob/main/LICENSE).
* The license for the compiled assets for on-device deployment can be found [here](https://github.com/facebookresearch/llama/blob/main/LICENSE)



## References
* [LLaMA: Open and Efficient Foundation Language Models](https://ai.meta.com/blog/meta-llama-3/)
* [Source Model Implementation](https://github.com/meta-llama/llama3/tree/main)



## Community
* Join [our AI Hub Slack community](https://qualcomm-ai-hub.slack.com/join/shared_invite/zt-2d5zsmas3-Sj0Q9TzslueCjS31eXG2UA#/shared-invite/email) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:[email protected]).

## 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