|
--- |
|
license: mit |
|
language: |
|
- en |
|
base_model: |
|
- microsoft/Phi-3.5-mini-instruct |
|
--- |
|
|
|
# Phi-3.5 Mini-Instruct Modification using MedIT-mesh Technique |
|
|
|
## Primary Use Cases: |
|
|
|
- Commercial use in environments requiring memory and compute constraints. |
|
- Use in latency-bound scenarios where accuracy is crucial. |
|
- Strong reasoning capabilities, especially for code, math, and logic applications. |
|
|
|
## Model Description: |
|
The Phi-3.5 Mini-Instruct modification is designed to accelerate research on language and multimodal models. It is a 3.8B parameter model optimized for commercial and research use in multiple languages. The MedIT-mesh technique provides improved memory and compute efficiency, making it suitable for environments with limited resources. |
|
|
|
## Use Case Considerations: |
|
|
|
When selecting use cases, developers should consider language models' limitations and evaluate accuracy, safety, and fairness before using them within a specific downstream application. |
|
Developers should be aware of applicable laws and regulations (e.g., privacy, trade compliance) relevant to their use case. |
|
It is essential to adhere to the license terms for the model being used. |
|
|
|
## Release Notes: |
|
|
|
An update over the June 2024 instruction-tuned Phi-3 Mini release based on user feedback. |
|
Additional post-training data was incorporated, leading to substantial gains in multilingual and multi-turn conversation quality, and reasoning capability. |
|
This release is expected to benefit most use cases, but users are encouraged to test in their particular AI applications. |
|
|
