Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pipeline_tag: text-generation
|
| 3 |
+
tags:
|
| 4 |
+
- NPU
|
| 5 |
+
---
|
| 6 |
+
# Phi-3.5-Mini
|
| 7 |
+
|
| 8 |
+
Run **Phi-3.5-Mini** optimized for **Qualcomm NPUs** with [nexaSDK](https://sdk.nexa.ai).
|
| 9 |
+
|
| 10 |
+
## Model Description
|
| 11 |
+
|
| 12 |
+
**Phi-3.5-Mini** is a \~3.8B-parameter instruction-tuned language model from Microsoft’s Phi family.
|
| 13 |
+
It’s designed to deliver strong reasoning and instruction-following quality within a compact footprint, making it ideal for **on-device** and **latency-sensitive** applications. This Turbo build uses Nexa’s Qualcomm NPU path for faster inference and higher throughput while preserving model quality.
|
| 14 |
+
|
| 15 |
+
## Features
|
| 16 |
+
|
| 17 |
+
* **Lightweight yet capable**: strong performance with small memory and compute budgets.
|
| 18 |
+
* **Conversational AI**: context-aware dialogue for assistants and agents.
|
| 19 |
+
* **Content generation**: drafting, completion, summarization, code comments, and more.
|
| 20 |
+
* **Reasoning & analysis**: math/logic step-by-step problem solving.
|
| 21 |
+
* **Multilingual**: supports understanding and generation across multiple languages.
|
| 22 |
+
* **Customizable**: fine-tune or apply adapters for domain-specific use.
|
| 23 |
+
|
| 24 |
+
## Use Cases
|
| 25 |
+
|
| 26 |
+
* Personal and enterprise chatbots
|
| 27 |
+
* On-device AI applications and offline assistants
|
| 28 |
+
* Document/report/email summarization
|
| 29 |
+
* Education and tutoring tools
|
| 30 |
+
* Vertical solutions (e.g., healthcare, finance, legal), with proper guardrails
|
| 31 |
+
|
| 32 |
+
## Inputs and Outputs
|
| 33 |
+
|
| 34 |
+
**Input**:
|
| 35 |
+
|
| 36 |
+
* Text prompts or conversation history (tokenized input sequences).
|
| 37 |
+
|
| 38 |
+
**Output**:
|
| 39 |
+
|
| 40 |
+
* Generated text: responses, explanations, or creative content.
|
| 41 |
+
* Optionally: raw logits/probabilities for advanced downstream tasks.
|
| 42 |
+
|
| 43 |
+
## License
|
| 44 |
+
|
| 45 |
+
* This model is released under the **Creative Commons Attribution–NonCommercial 4.0 (CC BY-NC 4.0)** license.
|
| 46 |
+
* Non-commercial use, modification, and redistribution are permitted with attribution.
|
| 47 |
+
* For commercial licensing, please contact **dev@nexa.ai**.
|
| 48 |
+
|
| 49 |
+
## References
|
| 50 |
+
|
| 51 |
+
* [Microsoft – Phi Models](https://www.microsoft.com/en-us/research/project/phi-3)
|
| 52 |
+
* [Hugging Face Model Card (Phi-3.5-Mini-Instruct)](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)
|
| 53 |
+
* [Phi-3 Technical Report (blog/overview)](https://azure.microsoft.com/en-us/blog/introducing-phi-3)
|