Add first draft
091ffff4
Use appropriate gelu function
7e66228d
More improvements
d3410104
More improvements
432ec566
More improvements
3d700aec
Convert checkpoint
61e98d1a
More improvements
4cfb533c
Improve docs, remove print statements
7bd5816e
More improvements
85e8efa8
Add link
5017a6a4
remove unused masking function
f07af7c2
begin tokenizer
64960412
do_lower_case
fcd729f5
debug
3cc8b4f5
set split_special_tokens=True
9e02a78d
Remove script
13f3f814
Fix style
c3818386
Fix rebase
82cd61b9
Use same design as CLIP
35fb3dcc
Add fast tokenizer
eb8748bd
Add SiglipTokenizer to init, remove extra_ids
59a616c6
Improve conversion script
da43b79f
Use smaller inputs in conversion script
9bf7faf5
Update conversion script
d40ed625
More improvements
59c2ed68
Add processor to conversion script
0d8e4b25
Add tests
5f982d23
Remove print statements
ed32e1fe
Add tokenizer tests
93c2eecd
Fix more tests
40bb3c03
More improvements related to weight initialization
e71278f8
More improvements
a6dcdbf1
Make more tests pass
b65ce001
More improvements
b210f0db
More improvements
3eaf473d
Add copied from
8de18230
Add canonicalize_text
6e44b7f2
Enable fast tokenizer tests
dd94da35
More improvements
1f1dc25b
Fix most slow tokenizer tests
da5a0d9b
Address comments
ec0e0371
Fix style
5b1dab16
Remove script
e2e6dc9a
xenova
requested changes
on 2023-12-04
xenova
commented
on 2023-12-05
Address some comments
bf0e06df
Add copied from to tests
ca53b4e4
Add more copied from
e4c991d6
Add more copied from
28ef8637
Add more copied from
d3d53630
Remove is_flax_available
e6c3d304
Fix merge
8d099d18
More updates
d20a7bda
Address comment
5c403a55
Remove SiglipTokenizerFast for now
395ec005
Add caching
96ddb2f4
Remove umt5 test
1cb8436e
Add canonicalize_text inside _tokenize, thanks Arthur
f6b8b1e4
Fix image processor tests
746297b7
Skip tests which are not applicable
5390e870
Skip test_initialization
0b68b5f6
More improvements
e6045e74
Compare pixel values
9e25c117
Fix doc tests, add integration test
0fe9ddd7
Add do_normalize
c7ae7b00
Remove causal mask and leverage ignore copy
3a7c4734
Fix attention_mask
60826769
Fix remaining tests
35d35a9d
Fix dummies
b4b129c6
Rename temperature and bias
30d8b5ef
Address comments
fe4a5442
Add copied from to tokenizer tests
a159f930
Add SiglipVisionModel to auto mapping
bfa7013f
Add copied from to image processor tests
888fcb8e
Improve doc
c78ab3ef
Remove SiglipVisionModel from index
d5bafa82
Address comments
c161c49f
Improve docs
446b1502
Simplify config
20de3094
Add first draft
c5590c9b
Make it like mistral
01738d26
More improvements
ae59c65e
Fix attention_mask
10479dac
Fix output_attentions
9a74702a
Add note in docs
70e49be7
Convert multilingual model
28bf56bd
Convert large checkpoint
a5258fd7
Convert more checkpoints
58f46ffb
Add pipeline support, correct image_mean and image_std
41bb11d7
Use padding=max_length by default
58937fb4
Make processor like llava
15dc528a
Add code snippet
3764b478
Convert more checkpoints
5ab84a3e
Set keep_punctuation_string=None as in OpenCLIP
985a9227
Set normalized=False for special tokens
ffdcfc5d
Fix doc test
f678d4cd
Update integration test
e8b82303
Add figure
ed40ec5f
Update organization
8c83814b
Happy new year
4a0e70cf
Use AutoModel everywhere
5e3df736
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub