Hugging Face's logo

zer0int
/
CLIP-GmP-ViT-L-14

Zero-Shot Image Classification
Transformers
Safetensors
clip
Inference Endpoints
Model card Files Community
5
zer0int commited on
Commit
b4aee98
•
1 Parent(s): 0518f9d

Greatly improved TEXT + Detail (as CLIP-L for Flux.1)

Browse files

High-Temperature Loss & tinkering, smaller modality gap. Otherwise based on GmP with label-smoothing (same as previous named "BEST" model), very similar ~91% ImageNet/ObjectNet and VOC-2007-multilabel accuracy.

Files changed (4) hide show
  1. ViT-L-14-TEXT-detail-improved-hiT-GmP-HF.safetensors +3 -0
  2. ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors +3 -0
  3. ViT-L-14-TEXT-detail-improved-hiT-GmP-pickle-OpenAI.pt +3 -0
  4. ViT-L-14-TEXT-detail-improved-hiT-GmP-state_dict.pt +3 -0
ViT-L-14-TEXT-detail-improved-hiT-GmP-HF.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39e79c916feca4ddf546d9fe923e664714b59ea61074f7228037d17c302f3d17
3
+ size 931448048
ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eddcf5e2e2a88550aab21e04895e8ce4b2977e6c11dbbc7675b07927a3cda0dd
3
+ size 323409740
ViT-L-14-TEXT-detail-improved-hiT-GmP-pickle-OpenAI.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5fe7fbd3b536404f5bb609b86dd6042b9e7f8e072b362fb1d8d683d73a32f3f
3
+ size 932296502
ViT-L-14-TEXT-detail-improved-hiT-GmP-state_dict.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14d8c7577ae4e2a79e355a21488c5c22c810ef663d062996de352b8a7bcf4318
3
+ size 932235192