Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
# moea-compression checkpoints
|
| 6 |
+
|
| 7 |
+
Pretrained ResNet-18 checkpoints for [moea-compression](https://github.com/rkhosrowshahli/moea-compression),
|
| 8 |
+
a multi-objective evolutionary search for uniform-binning weight quantization.
|
| 9 |
+
|
| 10 |
+
| file | dataset | architecture |
|
| 11 |
+
|---|---|---|
|
| 12 |
+
| `cifar10/resnet18/model_SA_best.pth.tar` | CIFAR-10 | ResNet-18 (3x3 stem, no maxpool) |
|
| 13 |
+
| `cifar100/resnet18/model_SA_best.pth.tar` | CIFAR-100 | ResNet-18 (3x3 stem, no maxpool) |
|
| 14 |
+
|
| 15 |
+
Each file is a `torch.save` dict with keys `state_dict`, `epoch`, `best_sa`.
|
| 16 |
+
Loaded automatically by `create_model()` in the moea-compression repo; no
|
| 17 |
+
manual download needed.
|