README.md

metadata

library_name: transformers
tags:
  - unsloth
  - trl
  - orpo
license: llama3

Model Card for ARC1

Self-instruction llama3-8b-instruct QLoRA fine-tune on generative abstraction & reasoning problem set.

Prompt Template

Instruction

<|begin_of_text|><|start_header_id|>system<|end_header_id|>
{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>
{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

Chat

<|begin_of_text|><|start_header_id|>system<|end_header_id|>
{{ system_prompt }}<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_1 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
{{ model_answer_1 }}<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_2 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>