Add recurrent gemma #30143
ArthurZucker
merged
97 commits into
main
from
add-recurrent-gemma
Merge pull request #9 from huggingface/update
e536f6a3
Fork.
c8dfc84e
RecurrentGemma initial commit.
a5821870
Updating __init__.py.
a6ba85f7
Minor modification to how we initialize the cache.
0ef8cca9
Reformat code to 4 spaces.
102770e5
Fixed the forward pass.
67710b01
Fixed the RecurrentGemmaForCausalLM
6427da15
Minor comment that we might not need attention_mask and output_attent…
0b57d56c
Now cache should work as well.
4056277f
Adding a temporary example to check whether the model generation works.
b3358c9a
Adding the tests and updating imports.
91c64f9d
Adding the example file missing in the previous commit.
19a008e3
First working example.
6a78aeff
Merge branch 'main' of github.com:huggingface/new-model-addition
ef8c0fb4
Removing .gitignore and reverting parts of __init__.
41d21dd7
Re-add .gitignore.
74c4b860
Addressing comments for configuration.
02400311
Move mask creation to `_prepare_inputs_for_generation`.
b8e4de48
First try at integration tests:
6085aa51
Transfoering between machines.
6d1f486f
Running normal tests.
211f06d7
Minor fix.
e276b26f
More fixes.
3887c140
Merge pull request #2 from botev/t1
66b3ca6e
Addressing more comments.
b62bb8bd
Minor fixes.
1644ee1b
first stab at cleanup
53f61e1c
more refactoring
9eb77d87
fix copies and else
94441f08
renaming and get init to work
c4dbe980
fix causal mask creation
23c54411
update
b68b253a
nit
951e2bd2
fix a hell lot of things
e0f3ca7a
updates
c272ef14
update conversion script
5ec60e5c
make all keys importable
8ccf919c
nits
55870df0
add auto mappings
3fe9f295
properly convert ffw_up and down
b819bc92
add scaling
ff473bf7
fix generations
91e4215c
for recurrent dtype
0a4bd574
update
213c313d
fix going beyong window
a903ab8a
fixup
928c1a1c
add missing files
bb4349b3
current updates to remove last einops
a490f855
finish modeling refactor
24abe59f
TADA
0f1b1beb
fix compile
5b6a2e62
fix most failing testt ? ?
57cc9e87
update tests
8d07eafb
refactor and update
68583497
update
5754cbbc
Merge remote-tracking branch 'botev/add-recurrent-gemma' into recurre…
9479fbc6
nits, fixup and update tests
f70554bd
more fixup
3d84644f
nits
6e0ccfc1
Merge branch 'main' of github.com:huggingface/transformers into recur…
a73ff678
fix imports
c700406e
test format
3bd8ed86
fixups
79527d57
nits
b81fe88e
tuple typing
589c2465
fix code quality
7dd3b476
add model card
370f7a64
fix doc
65d3bdd3
skip most generation tests
a360c969
nits
7dfb044e
style
6085f1de
doc fixes
ef05cc8c
fix pr and check_copies?
a34f905b
last nit
b0d2c981
oupsy
740a653a
Apply suggestions from code review
77963bad
update
cc80f2d0
Update src/transformers/models/recurrent_gemma/convert_recurrent_gemm…
308f3ead
Update tests/models/recurrent_gemma/test_modeling_recurrent_gemma.py
759bebe1
Update tests/models/recurrent_gemma/test_modeling_recurrent_gemma.py
4011bb04
Update tests/models/recurrent_gemma/test_modeling_recurrent_gemma.py
f153601b
Update tests/models/recurrent_gemma/test_modeling_recurrent_gemma.py
2505ff84
update based on review
a8849967
Merge branch 'add-recurrent-gemma' of github.com:huggingface/transfor…
9ad73e6c
doc nit
5a2ea8de
fix quality
c77723ef
quality
19121e42
fix slow test model path
c0af8880
update default dype
e20fb42d
ignore attributes that can be safely ignored in check config attributes
acf2c372
0lallalala come on
542af28c
save nit
2d26d92c
style
ea19cfbe
remove to dict update
ec6e0065
make sure we can also run in float16
9cb0c2dd
style
4b5ae335
proger
commented
on 2024-04-10
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub