Commit History
-
Update inference_tagger_standalone.py
Fix backbone state-dict loading: remap backbone.model.layer.* → backbone.layer.* The checkpoint stores the 32 transformer blocks under backbone.model.layer.N.* (HF-style, with an intermediate model wrapper), but DINOv3ViTH in this script declares them at backbone.layer.N.*. Combined with strict=False, assign=True in load_state_dict, all 608 block parameters (32 layers × 19 tensors per block) were silently failing to load — the backbone ran on default nn.Linear / nn.LayerNorm initializations while only the head loaded correctly. The only hint was a printed [Tagger] Missing keys (608): ['backbone.layer.0.layer_scale1', ...] line that was easy to miss, and the model produced plausible-looking but essentially random tag predictions, making it feel like undertraining. Confirmed by dumping tagger_proto.safetensors keys — they're all under backbone.model.layer.N.* and the head is a single projection.weight of shape (74625, 6400). Changes: Strip the intermediate model. segment from backbone k…3136b21
-
tagger_low_rank_head
93a0200
-
tagger_low_rank_head
49f290c
-
tagger_low_rank_head
6c1b3f6
-
tagger_low_rank_head
e6efd89
-
tagger_low_rank_head
5ccf57f
-
tagger_low_rank_head
81f1a0d
-
tagger_low_rank_head
93c389c
-
tagger_low_rank_head
e9b42c5
-
tagger_low_rank_head
b7734bc
-
tagger_low_rank_head
07db30b
-
tagger_low_rank_head
29a191c
-
tagger_low_rank_head
96897a8 -
Update README.md
7d3fa21 -
Upload tagger_vocab_with_categories.json
6867339 -
Update tagger_ui_server.py
ccc01da -
Update tagger_ui/templates/index.html
45887e5
-
Update README.md
74f5ca2 -
Update README.md
ee71c01 -
Update README.md
49e3315 -
Update README.md
e59eadd -
tagger_low_rank_head
5ab7ede -
Upload tagger_ui/templates/index.html
4e7ac21 -
Upload tagger_ui_server.py
bc0d1c8 -
Upload inference_tagger_standalone.py
3625530
-
Upload tagger_vocab.json
d1eb9c4 -
tagger_low_rank_head
a81c6b8 -
initial commit
39a186d