transformers
Add EfficientLoFTR model
#36355
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
81
Changes
View On
GitHub
Add EfficientLoFTR model
#36355
qubvel
merged 81 commits into
huggingface:main
from
sbucaille:add_eloftr
qubvel
requested a review
from
qubvel
1 year ago
qubvel
added
New model
qubvel
added
Vision
qubvel
commented on 2025-03-03
sbucaille
force pushed
from
8eada84c
to
8d6a0484
1 year ago
sbucaille
force pushed
from
8098b16c
to
a60f4cdb
1 year ago
qubvel
requested a review
from
qubvel
1 year ago
qubvel
commented on 2025-03-13
sbucaille
force pushed
from
e7d0b8e4
to
5acfdb58
1 year ago
qubvel
requested a review
from
qubvel
1 year ago
qubvel
commented on 2025-04-01
sbucaille
force pushed
from
9d130fa4
to
b5117b6a
1 year ago
sbucaille
force pushed
from
4ec83cc0
to
fdd377a0
1 year ago
qubvel
requested a review
from
qubvel
1 year ago
sbucaille
force pushed
from
9b154f2b
to
ab0a051d
1 year ago
initial commit
93256f2a
Apply suggestions from code review
e969a1e6
fix: various typos, typehints, refactors from suggestions
492dccff
fix: fine_matching method
7dd04830
Added EfficientLoFTRModel and AutoModelForKeypointMatching class
4c748e87
fix: got rid of compilation breaking instructions
62d95a59
docs: added todo for plot
560bfa2e
fix: used correct hub repo
8c8d4de5
docs: added comments
e29bb261
fix: run modular
ed6e0838
doc: added PyTorch badge
e64b4c50
fix: model repo typo in config
14dcd0e7
fix: make modular
8cafdaba
fix: removed mask values from outputs
5ccfc509
feat: added plot_keypoint_matching to EfficientLoFTRImageProcessor
172729e3
feat: added SuperGlueForKeypointMatching to AutoModelForKeypointMatch…
d20ccfc7
fix: reformat
19c08954
refactor: renamed aggregation_sizes config parameter into q, kv aggre…
0f961f1b
doc: added q, kv aggregation kernel size and stride doc to config
5ee4145a
refactor: converted efficientloftr implementation from modular to cop…
bec741ae
tests: overwrote batching_equivalence for "keypoints" specific tests
f47ea90e
fix: changed EfficientLoFTRConfig import in test_modeling_rope_utils
82929c01
fix: make fix-copies
4901d3b2
fix: make style
39cd0ea2
fix: update rope function to make meta tests pass
0e632900
fix: rename plot_keypoint_matching to visualize_output for clarity
0f9dd7c8
refactor: optimize image pair processing by removing redundant target…
58d728b8
feat: add EfficientLoFTRImageProcessor to image processor mapping
97dfa2e6
refactor: removed logger and updated attention forward
22226c34
refactor: added auto_docstring and can_return_tuple decorators
0c924b20
refactor: update type imports
e729ba3f
refactor: update type hints from List/Dict to list/dict for consistency
e9b85e41
refactor: update MODEL_MAPPING_NAMES and __all__ to include LightGlue…
3d9669ab
sbucaille
force pushed
from
49419e08
to
3d9669ab
1 year ago
fix: change type hint for size parameter in EfficientLoFTRImageProces…
c7fdaf91
fix typing
4ecf5c90
Merge branch 'main' into pr/sbucaille/36355
39192c46
fix some typing issues
f47756be
nit
4cfde9ac
a few more typehint fixes
f48146da
Remove output_attentions and output_hidden_states from modeling code
b6e2a82f
else -> elif to support efficientloftr
8fb411f7
nit
4508db2b
qubvel
commented on 2025-07-14
tests: added EfficientLoFTR image processor tests
4c5a4535
refactor: reorder functions
9feb4f69
chore: update copyright year in EfficientLoFTR test file
3aa9aff8
Use default rope
08d6615f
Add docs
890d6dc4
Update visualization method
1bbfec90
fix doc order
a614057b
remove 2d rope test
a7183a0d
qubvel
approved these changes on 2025-07-15
Update src/transformers/models/efficientloftr/modeling_efficientloftr.py
f6dc36d8
fix docs
f3f07e9b
qubvel
commented on 2025-07-15
Update src/transformers/models/efficientloftr/image_processing_effici…
bbda03fd
ArthurZucker
commented on 2025-07-15
update gradient
f59af0df
refactor: removed unused codepath
d842f63b
Add motivation to keep postprocessing in modeling code
28834dff
refactor: removed unnecessary variable declarations
c15edba1
docs: use load_image from image_utils
3355fad5
refactor: moved stage in and out channels computation to configuration
d41fb6be
refactor: set an intermediate_size parameter to be more explicit
09463ad5
refactor: removed all mentions of attention masks as they are not used
ec4d65ac
refactor: moved position_embeddings to be computed once in the model …
6c8b6fd1
qubvel
commented on 2025-07-16
refactor: removed unnecessary hidden expansion parameter from config
71fb8c2b
refactor: removed completely hidden expansions
b314d98f
refactor: removed position embeddings slice function
cc70a373
tests: fixed broken tests because of previous commit
7878d1a6
qubvel
requested a review
from
qubvel
1 year ago
fix is_grayscale typehint
1fd91042
not refactoring
de6f5d0f
not renaming
4e4df87d
move h/w to embeddings class
c69fde30
Precompute embeddings in init
abecd173
Merge branch 'main' into add_eloftr
3fb3ba0c
fix: replaced cuda device in convert script to accelerate device
50ee9a81
fix: replaced stevenbucaille repo to zju-community
128fd425
Merge branch 'main' into add_eloftr
5662c54e
Remove accelerator.device from conversion script
b1935766
ArthurZucker
approved these changes on 2025-07-21
refactor: moved parameter computation in configuration instead of fig…
eb80048e
fix: removed unused attributes in configuration
525db053
fix: missing self
1552467b
fix: refactoring and tests
bace3c95
fix: make style
d29bb512
qubvel
approved these changes on 2025-07-22
qubvel
merged
a88ea9cb
into main
1 year ago
sbucaille
deleted the add_eloftr branch
1 year ago
Login to write a write a comment.
Login via GitHub
Reviewers
ArthurZucker
qubvel
Assignees
No one assigned
Labels
New model
Vision
Milestone
No milestone
Login to write a write a comment.
Login via GitHub