Phi-3
Phi-3Model SizePerformancePulling Phi-3Using Phi-3Running Phi-3ChattingEnding the conversationReferences
Demo Environment
Development Board: Jetson Nano
SD (TF) Card: 64GB
Recommended for models with 4 bytes or fewer parameters
Phi-3 is a powerful, cost-effective Small Language Model (SLM) from Microsoft, outperforming models of the same size and larger across various language, reasoning, encoding, and math benchmarks.
Model Location
xxxxxxxxxx/usr/share/ollama/.ollama/models
Model Size
| Model | Parameters |
|---|---|
| Phi-3 (Mini) | 3.8 Bytes |
| Phi-3 (Medium) | 14 Bytes |
xxxxxxxxxxJetson Nano: Tested using a Phi-3 model with 3.8 Bytes!
Performance
Pulling Phi-3
Using the pull command will automatically pull the model from the Ollama model repository:
xxxxxxxxxxollama pull phi3:3.8b
Using Phi-3
Running Phi-3
If the system does not have a running model, the system will automatically pull the Phi-3 3.8B model and run it:
xxxxxxxxxxollama run phi3:3.8b
Chatting
xxxxxxxxxxTell me a bedtime story
Response time depends on your hardware configuration, so please be patient!
Ending the conversation
Use the Ctrl+d shortcut or /bye to end the conversation!
References
Ollama
Official website: https://ollama.com/
GitHub: https://github.com/ollama/ollama
Phi-3
Ollama corresponding model: https://ollama.com/library/phi3