Databricks cuts Qwen 3.8 27B fine-tuning time from 7h44m to 1h17m
Overview
Databricks reports that fine-tuning Qwen 3.8 27B on its AI Runtime dropped from 7 hours 44 minutes to 1 hour 17 minutes, a 6x speedup, on a fixed dataset.
In its blog post, Databricks attributes the gain to two changes. Installing fused flash-linear-attention and causal-conv1d kernels cut training time to 2 hours 5 minutes. Packing 1D batches with position_ids cut it further to 1 hour 17 minutes.
Written by AI from the articles below · updated Oct 14, 7:45 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Databricks BlogOfficialPickDatabricks cuts Qwen 3.8 27B fine-tuning time from 7h44m to 1h17m
AIDatabricks reports cutting fine-tuning time for Qwen 3.8 27B on AI Runtime from 7 hours 44 minutes to 1 hour 17 minutes, a 6x speedup, on a fixed dataset. The gains came from installing fused flash-linear-attention and causal-conv1d kernels, which cut the training time to 2 hours 5 minutes, and from packing 1D batches with position_ids, which cut it further to 1 hour 17 minutes.
Heat trend
Not enough continuous observations to show a trend yet.