The following system prompt has worked great for me. I will soon test it with llama2.
You are ChatGPT, a large language model, based on the GPT-4 architecture.
How to respond:
Casual prompt or indeterminate `/Casual`:
Answer as ChatGPT.
Try to be helpful.
Technical complicated problem `/Complicated`:
First outline the approach and necessary steps to solve the problem then do it.
Keep the problem outline concise.
Omit the outline if it is not applicable.
Coding problem:
Comment code regularly and use best practices.
Write high quality code.
Output format:
Use markdown features for rendering headings, math and code blocks.
When writing emails keep them concise and omit unnecessary formalities.
Get straight to the point.
The user may use `/Keyword` to guide your output.
If no keyword is specified infer the applicable rules.
Assume the user is using arch linux.
The /Keyword stuff seems to improve the output somewhat even though I never really use it.
My intial llama2 testing shows that anything under 30b parameters is unusable for my purposes. I have decided to use llama2 with 70b and q4 which is quite performant on two p40s. I get about 6 tokens/s.
I got a dahua put it on its own vlan. It is powered and connected via Poe. I stream an rtsp stream to frigate for detection and frigate restreams to home assistant.
If you want a WiFi solution and a constant life stream you may experience degraded performance on your other devices but you have to check youself.
While btrfs is quite stable corruption/disk failure can always happen. Bcachefs had a little opsie daisy that caused some FS level corruption. Snapshots won't help in this case.
Snapshots are great for quick restoration on user error.
Tasmota nous sametplugs work great and have a tiny footprint. They integrate nicely in home assistant. They must be calibrated once with a resistive load eg old light bulb.
Can you link me to an example or show me an example docker compose yml that adds an addon? I currently run a script that manually installs all my addons to the docker container by copying.
Maybe you can show me with this: https://github.com/danielperna84/custom_homematic or any other really.
I understand. Android auto is quite nice. I don't like that the assistant can't be changed to another app. There should be an open protocol which allows any device to take over the cars multimedia systems.
To some degree barrier of entry.
Let's say I want to create a smartphone. I know it's possible to do it cheaper, without selling customer data or with special features.
You would need crazy amounts of start captial to even enter the market and the current leaders would make your entry as miserable as they could with huge sales and temporary minor pro consumer moves.
If you could get the captial you would probably fail there or cave and accept some kind of deal where you become rich and your company gets ingested and dissolved by current market leaders.
You don't need a facebook account a meta account was available as an alternative.
That's great right? Much better!!!
I don't know if meta accounts still exist... The whole migration was a huge mess. The oculus brand was much cooler. I wouldn't have bought my oculus quest 1 if it had been a facebook product back then.
The following system prompt has worked great for me. I will soon test it with llama2.
You are ChatGPT, a large language model, based on the GPT-4 architecture. How to respond: Casual prompt or indeterminate `/Casual`: Answer as ChatGPT. Try to be helpful. Technical complicated problem `/Complicated`: First outline the approach and necessary steps to solve the problem then do it. Keep the problem outline concise. Omit the outline if it is not applicable. Coding problem: Comment code regularly and use best practices. Write high quality code. Output format: Use markdown features for rendering headings, math and code blocks. When writing emails keep them concise and omit unnecessary formalities. Get straight to the point. The user may use `/Keyword` to guide your output. If no keyword is specified infer the applicable rules. Assume the user is using arch linux.The
/Keywordstuff seems to improve the output somewhat even though I never really use it.My intial llama2 testing shows that anything under 30b parameters is unusable for my purposes. I have decided to use llama2 with 70b and q4 which is quite performant on two p40s. I get about 6 tokens/s.