sbucaille
marked this pull request as draft 272 days ago
sbucaille
force pushed
from
6a4faac2
to
15aa78c4
269 days ago
sbucaille
marked this pull request as ready for review 269 days ago
sbucaille
changed the title [WIP] Add LWDetr model Add LWDetr model 269 days ago
sbucaille
force pushed
from
f198fbc1
to
03e3d834
265 days ago
sbucaille
force pushed
from
0c1f23b1
to
f35fb75d
243 days ago
sbucaille
force pushed
from
f35fb75d
to
c4f2893b
240 days ago
sbucaille
force pushed
from
814a1916
to
284b51ce
208 days ago
sbucaille
force pushed
from
284b51ce
to
47c1cb46
208 days ago
feat: add LWDetr model
23631b17
fix: changed LwDetrVit base classes from VitDet to ViT
0eb69e6e
tests: added tests for LWDetr
12fa5e23
refactor: fix all issues and created docs
aceb10c3
tests: added missing lw_detr_vit tests
bd48206b
docs: add lwdetr docs
06c7d70c
fix: fixed implementation error and associated tests
a2ef8c34
chore: removed testing lib in imports
9faeaee8
refactor: replace LwDetrImageProcessor with DeformableDetrImageProcessor
0fad3404
refactor: remove two-stage detection and bounding box reparameterizat…
a89f8f29
refactor: rename LwDetrCSPRepLayer to LwDetrC2FLayer
e942891e
refactor: introduce LwDetrMLP for feedforward layers in decoder
f9e96317
refactor: replace build_position_encoding with LwDetrSinePositionEmbe…
9cf545d6
refactor: remove use_cae parameter and related logic from configurati…
199b2bc6
refactor: remove unused variables and simplify certain instructions
ab6096f7
refactor: removed unnecessary one line instruction method with_pos_embed
b95de124
refactor: use llama attention formatting for hidden shape
0af67c2d
docs: add comments about group detr
15625a50
fix: removed wrong sigmoid and fixed init for class_embed
22d66d2f
refactor: removed unused positional embeddings classes and weights fr…
5b7f6570
chore: removed unused import
b075292d
chore: make style and repo-consistency after positional embeddings re…
d863a68a
refactor: removed unused drop path rate
d6fdd91b
fix: ingest latest changes from rebase
13ad4a8a
fix: attn_implementation setter
8147b456
fix: is causal set to False
25fbaab1
refactor: renamed ffn to mlp and moved layer norm out of mlp
d5b24a6a
fix: check model inputs
fa6deed9
fix: moved super init call in LwDetrConfig
75b3f1fc
fix: super class in GradientCheckpointingLayer
f998abb9
fix: replaced RTDetr occurences by LwDetr in test modeling file
e164b8a9
refactor: removed head_mask from LwDetrViT
05afaa7c
docs: added release date in docs
ff488212
fix: added missing attention mask argument
c627a2a9
chore: make style & repo-consistency
755c5b84
fix: ensure tensor dtype consistency in loss calculations and test cases
8b72816d
docs: fixed model release date
2b9ebffa
refactor: removed unnecessary module cloning
6e4f5839
tests: added missing _prepare_for_class method and removed batching_e…
97c1d37a
tests: added xlarge integration test
abae3759
chore: added lw_detr reference in image processing auto
e037e63a
chore: removed unnecessary properties from LwDetrConfig
9deff21d
fix: fix for latest main changes
be4fc9f1
fix: apply modular changes from mail
7556efca
docs: update model doc and docstrings
77a94e7b
sbucaille
force pushed
from
e462c9ee
to
77a94e7b
197 days ago
fix: style
fec5db9c
fix: update output values in convert script
4cdc807b
stevhliu
approved these changes
on 2025-12-04
feat: added proper last_hidden_states in LwDetrDecoderOutput and sepa…
df6f2eda
fix: guard accelerate imports
138b009c
fix: removed LWDetrConfig attribute map and changed LwDetrAttention i…
d71dbb81
fix: parameterize amap based on config
f9e60b42
fix: remove redundant decorator
635f5279
chore: moved LwDetrViT to LwDetr single modular file
eeac74a3
fix: remove unnecessary attribute_map in LwDetrViT
514536e9
chore: simplified LwDetr modules methods with proper hidden_states re…
e82f6d5f
fix: replaced hardcoded value by variable
13e9aa36
tests: added VitDet and attention tests
082715b1
fix: modular conversion
74c47a73
tests: moved LwDetrViT tests to test_modeling_lw_detr file
865739f6
docs: add lwdetr advances in docs
dec88cd6
stevhliu
approved these changes
on 2025-12-10
refactor: removed arguments to classes as much as possible and rely o…
b7821a38
Merge branch 'main' into add_lw_detr
c9809f8e
reapply style, remove LlamaAttention inheritance to remove decorator
ad93a7b5
chore: updated licence and year
e5a0446d
fix: removed torch.nn.functional from modular
003d63d6
docs: removed redundant docstring arguments covered by autodocstring …
99490ca4
refactor: removed backbone api statements
11727aba
fix: added back num_key_value_groups in LwDetrAttention
8b8feb89
chore: removed unnecessary copied from statement
c68e713a
chore: moved LwDetrViT modules above LwDetr modules
d0cfb7ab
tests: removed unnecessary overwrite and “test_” attributes
c03d5ee7
docs: added missing docs
3342ea88
Merge remote-tracking branch 'upstream/main' into add_lw_detr
b01b7c54
style: remove unnecessary parentheses
8e2753e0
docs: added back logits docstring
06ac9fb9
docs: added docs dates
b150f230
Merge branch 'main' into add_lw_detr
6f893885
style details
8a7818f4
unessecary utf8
e5c20a0a
might as well skip all config checks
c745f904
embeddings are large, increase model_split_percents
098fb4df
fix device issue
93ba55ac
update logits
aabdb76d
set device in expectations
0493e06a
add to toctree
1312458d
sbucaille
deleted the add_lw_detr branch 157 days ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub