metadata
language:
- en
license: apache-2.0
tags:
- text-generation-inference
- transformers
- unsloth
- mistral
- trl
base_model: unsloth/mistral-7b-bnb-4bit
"""
- an attempt at making a working "program" inside an llm thru fine tuning. kinda works kinda doesn't
- commands are:
- start/start box/box
- up/down/left/right
- heres an example convo:
- user: "start box"
- ai: " - 12345 - _____ - 1|X....| - 2|.....| - 3|.....| - 4|.....| - 5|.....| - _____ -
- start"
- user: "right"
- ai: "right
- - 12345 - _____ - 1|.X...| - 2|.....| - 3|.....| - 4|.....| - 5|.....| - _____ -
"
- user: "down"
- ai: "down
- - 12345 - _____ - 1|.....| - 2|.X...| - 3|.....| - 4|.....| - 5|.....| - _____ -
"
- user: "right"
- ai: "right
- - 12345 - _____ - 1|.....| - 2|..X..| - 3|.....| - 4|.....| - 5|.....| - _____ -
"
- user: "up"
- ai: "up
- - 12345 - _____ - 1|..X..| - 2|.....| - 3|.....| - 4|.....| - 5|.....| - _____ -
"
- user: "left"
- ai: "left
- error: unable to move left from this position
- - 12345 - _____ - 1|..X..| - 2|.....| - 3|.....| - 4|.....| - 5|.....| - _____ -
"
- after this it stopped working but you get the gist. imagine playing doom thrun an llm lol
- """
Uploaded model
- Developed by: liminerity
- License: apache-2.0
- Finetuned from model : unsloth/mistral-7b-bnb-4bit
This mistral model was trained 2x faster with Unsloth and Huggingface's TRL library.