This model was trained as part of a series of experiments testing the performance of pure DPO vs SFT vs ORPO, all supported by Unsloth/Huggingface TRL.
Note: Completely broken. Do not use.
Benchmarks
Average 59.52
ARC 59.47
HellaSwag 82.42
MMLU 62.21
TruthfulQA 40.01
Winogrande 78.3
GSM8K 34.72
Training Details
Duration: ~10-12 hours on one Kaggle T4 with Unsloth
Model: https://huggingface.co/unsloth/mistral-7b-v0.2-bnb-4bit
Dataset: https://huggingface.co/datasets/argilla/dpo-mix-7k
Rank: 8
Alpha: 16
Learning rate: 5e-6
Beta: 0.1
Batch size: 8
Epochs: 1
Learning rate scheduler: Linear
Prompt Format: You are a helpful assistant.<s>[INST] PROMPT [/INST]RESPONSE</s> (The start token <s> must be added manually and not automatically)
WanDB Reports
