Phi 3 Mini 128K Instruct is an open-source language model by inventbot. Features: 3.8b LLM, VRAM: 7.7GB, Context: 128K, License: mit, Instruction-Based, LLM Explorer Score: 0.23, ELO: 1129.
Commercial and research use in English, Memory/compute constrained environments, Latency bound scenarios, Strong reasoning (code, math, logic)
Primary Use Cases:
Generative AI powered features
Limitations:
Not specifically designed for all downstream purposes
Considerations:
Evaluate and mitigate for accuracy, safety, and fairness before use in high-risk scenarios.
Additional Notes
The model uses flash attention which requires certain types of GPU hardware.
Supported Languages
en (Primary language supported)
Training Details
Data Sources:
Publicly available documents filtered rigorously for quality, Selected high-quality educational data, Code, Newly created synthetic "textbook-like" data, High quality chat format supervised data covering various topics
Data Volume:
3.3T tokens
Methodology:
Supervised fine-tuning and Direct Preference Optimization
Context Length:
128000
Training Time:
7 days
Hardware Used:
512 H100-80G GPUs
Model Architecture:
Dense decoder-only Transformer model
Safety Evaluation
Methodologies:
Safety post-training
Risk Categories:
Misinformation, Bias, Inappropriate or Offensive Content
Ethical Considerations:
Developers should apply responsible AI best practices and are responsible for ensuring compliance with relevant laws.
Responsible Ai Considerations
Fairness:
Models can potentially behave in ways that are unfair.
Transparency:
Models can generate nonsensical content or fabricate content.
Accountability:
Developers should assess suitability for high-risk scenarios.
Mitigation Strategies:
Using Retrieval Augmented Generation (RAG) and feedback mechanisms.
Input Output
Input Format:
Chat format
Accepted Modalities:
text
Output Format:
Generated text in response to input
Performance Tips:
Use the chat format for better results with the provided template.