# How Lyft uses Weights & Biases

**4.3x**: “we speeded up EEF training by 4.3x and also improved model accuracies”

As published on [wandb.ai](https://wandb.ai/site/customers/lyfts-high-capacity-end-to-end-camera-lidar-fusion-for-3d-detection/), October 2023. Captured by usedby on 2026-10-08.

- Company: [Lyft](https://www.usedby.ai/companies/lyft.md)
- Tool: [Weights & Biases](https://www.usedby.ai/tools/weights-and-biases.md)
- Industry: [Mobility & Ride-hailing](https://www.usedby.ai/companies/industry/mobility-ride-hailing.md)
- Teams: Lyft Level 5

## What the story says

Lyft Level 5 used Weights & Biases, alongside its internal ML framework Jadoo, to run experiments on end-to-end camera-lidar fusion models for 3D detection in autonomous vehicles. The experiments compared fusion strategies, model capacity, data augmentation, and training speed optimizations.

Summary written by usedby from the source page, in English. The figures are those of Weights & Biases and Lyft, not ours.

> With the help from both image dropout and half-res images, we speeded up EEF training by 4.3x and also improved model accuracies.

- **1.2x**: “We used the automatic mixed-precision from PyTorch and further speeded up training by 1.2x, reduced GPU memory footprint without regressing model accuracies.”
- **2.3x**: “This reduced compute and data loading time and led to 2.3x faster training.”
- **2x**: “This further led to 2x faster training than using full-res images without much accuracy regression.”
- **1.5%**: “we observed better accuracies (shown below: +1.5% AP@0.5 for pedestrians).”

## What usedby checked

We compared the story with its live page on 2026-10-08.

- Checked: the figure 4.3x is printed word for word on the page, near the name of Lyft.
- Checked: the passage quoted above is copied word for word from the page, near the name of Lyft.
- Checked: each number in our summary is printed on the page.
- Checked: the publication date is read from the page’s own metadata, never guessed.
- Checked: Lyft uses Weights & Biases. Confirmed line. Latest check across sources: 2026-10-08.
- Not checked: the result itself. We quote it; we did not measure it.

---
Source: https://www.usedby.ai/case-studies/lyft-weights-and-biases · How we check: https://www.usedby.ai/methodology