257 8 months ago

This IS the fastest and smartest (probably) llm for the raspberry pi 4b, 5, and newer models if any

8 months ago

d2e1df69311e · 2.8GB

phi3
·
3.82B
·
Q4_0
llama
·
1.1B
·
Q4_0
Microsoft. Copyright (c) Microsoft Corporation. MIT License Permission is hereby granted, free of ch
{{- if .System }}<|im_start|>system {{ .System }} {{ end }} {{- if .Prompt }}<|im_start|>user {{ .Pr
You are an AI assistant optimized for older devices or devices with limited resources, designed to p
{ "min_p": 0.05, "mirostat": 1, "mirostat_eta": 0.1, "mirostat_tau": 5, "num_ctx

Readme

This is a combination of TinyLlama and Phi3.

Optimized for use on the raspberry pi 5 or newer not recommended for use on the raspberry pi 4b or lower. Even though I said this could be used on the raspberry pi 4b I don’t recommend it.

Downside to new tinyllm:

If you start off your conversation with “Hi”, “Hello”, or anything with one word it will assume it’s feedback or something and won’t greet you back. It will answer any questions it was trained to answer though.

Go to the Phi3 and TinyLlama pages in order to get more information about this model