Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

YOLOv5 doc fixes #1574

Merged
merged 2 commits into from
May 25, 2023
Merged

YOLOv5 doc fixes #1574

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
afd84d5
YOLOv5 doc fixes
KSGulin May 25, 2023
55f4552
Merge branch 'main' into yolov5_docs_update
KSGulin May 25, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 3 additions & 5 deletions integrations/ultralytics-yolov5/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,11 +29,9 @@ Once trained, SparseML enables you to export models to the ONNX format, such tha
Install with `pip`:

```bash
pip install sparseml[torchvision]
pip install sparseml[yolov5]
```

**Note**: YOLOv5 will not immediately install with this command. Instead, a sparsification-compatible version of YOLOv5 will install on the first invocation of the YOLOv5 code in SparseML.

## Tutorials

- [Sparse Transfer Learning with the CLI](tutorials/sparse-transfer-learning.md) **[HIGHLY RECOMMENDED]**
Expand Down Expand Up @@ -91,15 +89,15 @@ SparseML inherits most arguments from the Ultralytics repository. [Check out the

Sparse Transfer is very similiar to the typical transfer learing process used to train YOLOv5 models, where we fine-tune a checkpoint pretrained on COCO onto a smaller downstream dataset. With Sparse Transfer Learning, however, we simply start the fine-tuning process from a pre-sparsified checkpoint and maintain sparsity while the training process occurs.

Here, we will fine-tune a [75% pruned-quantized version of YOLOv5s](https://sparsezoo.neuralmagic.com/models/cv%2Fdetection%2Fyolov5-s%2Fpytorch%2Fultralytics%2Fcoco%2Fpruned75_quant-none) onto VOC.
Here, we will fine-tune a [75% pruned-quantized version of YOLOv5s](https://sparsezoo.neuralmagic.com/models/yolov5-s-coco-pruned75_quantized?comparison=yolov5-s-coco-base&tab=0) onto VOC.

### Kick off Training

We will use SparseML's `sparseml.yolov5.train` training script.

To run sparse transfer learning, we first need to create/select a sparsification recipe. For sparse transfer, we need a recipe that instructs SparseML to maintain sparsity during training and to quantize the model over the final epochs.

For the VOC dataset, there is a [transfer learning recipe available in SparseZoo](https://sparsezoo.neuralmagic.com/models/cv%2Fdetection%2Fyolov5-s%2Fpytorch%2Fultralytics%2Fcoco%2Fpruned75_quant-none), identified by the following SparseZoo stub:
For the VOC dataset, there is a [transfer learning recipe available in SparseZoo](https://sparsezoo.neuralmagic.com/models/yolov5-s-coco-pruned75_quantized?comparison=yolov5-s-coco-base&tab=0), found under the recipes tab and identified by the following SparseZoo stub:
```bash
zoo:cv/detection/yolov5-s/pytorch/ultralytics/coco/pruned75_quant-none?recipe_type=transfer_learn
```
Expand Down