Threshold Filtering Packing for Supervised Fine-Tuning: Training Related Samples within Packs
This addresses efficiency and performance issues in fine-tuning large language models, though it is an incremental improvement over existing packing methods.
The paper tackles the problem of cross-contamination and lack of learning signals in packing for supervised fine-tuning of autoregressive models by introducing Threshold Filtering Packing, which selects related samples within packs, resulting in improvements of up to 7% on GSM8K, 4% on HumanEval, and 15% on bias benchmarks.
Packing for Supervised Fine-Tuning (SFT) in autoregressive models involves concatenating data points of varying lengths until reaching the designed maximum length to facilitate GPU processing. However, randomly concatenating data points can lead to cross-contamination of sequences due to the significant difference in their subject matter. The mainstream approaches in SFT ensure that each token in the attention calculation phase only focuses on tokens within its own short sequence, without providing additional learning signals for the preceding context. To address these challenges, we introduce Threshold Filtering Packing (TFP), a method that selects samples with related context while maintaining sufficient diversity within the same pack. Our experiments show that TFP offers a simple-to-implement and scalable approach that significantly enhances SFT performance, with observed improvements of up to 7\% on GSM8K, 4\% on HumanEval. Furthermore, results from bias benchmark datasets highlight TFP's promising performance in improving fairness while also boosting prediction accuracy by 15\%.