Sanitize EMA checkpoint and align paper metrics

README.md

@@ -1,7 +1,5 @@
 ---
 license: mit
-library_name: timm
-pipeline_tag: image-classification
 tags:
 - image-classification
 - vision-transformer
@@ -11,13 +9,12 @@ datasets:
 - ILSVRC/imagenet-1k
 ---
 
-# progresvit-deit-s-192-240-imagenet1k
+# ProgResViT DeiT-S (192 → 240)
 
-EMA checkpoint for the DeiT-S ProgResViT model trained on ImageNet-1K.
+EMA inference checkpoint for the DeiT-S ProgResViT model trained on ImageNet-1K.
 
-- Resolution schedule: `192 -> 240`
-- Training: Standard
-- Full-path top-1 accuracy: 82.202%
+- Training: standard
+- Paper-reported full-path top-1 accuracy: 82.206%
 - Full-path compute: 6.267 GMACs
 
 ## Download
@@ -31,4 +28,8 @@ checkpoint = hf_hub_download(
 )
 ```
 
-Training and evaluation code: https://github.com/ds-kiel/ProgResViT
+The file contains EMA weights and public architecture metadata only. Use it with the [ProgResViT repository](https://github.com/ds-kiel/ProgResViT):
+
+```bash
+DATA_DIR=/path/to/imagenet CONFIG=192_240 bash scripts/validate.sh
+```

progresvit_192_240.pth.tar

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8d40db265b8da68a9cf16cd3c20ae29cd40ab03d02c02e939b9619a8ec6ca130
-size 396242239
+oid sha256:fab254d778b2c47b2fb6909169205564f7696336f86e3db21309f880fb25617f
+size 99032753