Skip to content
This repository has been archived by the owner on Jan 21, 2024. It is now read-only.

v.0.1.0
Compare
Choose a tag to compare
@dabumana dabumana released this 25 Jan 02:20
· 246 commits to main since this release
v.0.1.0
dcab526
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

⚠️ Change Log:

Test all the available models for code/text/insert/similarity

Modes:

Training mode: Prepare your own sets based on the interaction

Edit mode: First input will be contextual the second one instructional

Conversational mode: A friendly experience with the model

Advanced parameters like:

  • Temperature
  • Topp
  • Penalty
  • Frequency penalty
  • Max tokens
  • Engine
  • Multiple results and probabilities
  • Detailed log according to UTC
  • Export JSON files for training purposes
  • Export conversations on TXT

General parameters:

Mode: Modify the actual mode, select between (TEXT/EDIT/CODE)
Engine: Modify the model that you want to test
Results: Modify the amount of results displayed for each prompt
Probabilities: According to your setup of the temperature and topp, probably you will need to use this field to populate a more accurate response according to the possibilities of results
Temperature: If you are working with temperature, try to keep the topp in a higher values than temperature
Topp: Applies the same concept as temperature, when you are modifying this value, you need to apply a higher value for temperature
Penalty: Penalty applied to the characters an redundancy in a result completion
Frequency Penalty: Establish the frequency of the penalty threshold defined

Modes:

Training Mode: Prepare your fine tunning data based on the completion response, can be exported to a JSON file for further use.
Edit Mode: Use Edit mode for all the requests
Press New Conversation and select Edit mode the first request will be for a completion endpoint the second based on the first request will continue editing the content in the parameters that you ask.
Conversational Mode: Use conversational AI mode request for a friendly interaction.

Assets 2