Add kyutai stt #38909
LysandreJik
merged
107 commits into
huggingface:main
from
eustlb:add-kyutai-stt
first draft
936792a4
Merge branch 'huggingface:main' into moshi-asr
111f3ea4
cleaner version
536e55d6
Merge branch 'huggingface:main' into moshi-asr
53f97436
udpate tests + modeling
22be12c5
add tests
ca508a04
init
f83bbb00
udpate test_modeling_common
ff510f75
fix tests
af3ff35f
csm Processor draft
efd67360
convertion update
c132465e
mimi cache padding convolutions draft
9966e568
mimi streaming udpates
4c1feb39
update mimi padding cache test
72289efe
udpate cache padding mimi test
8023ebdd
make style mimi
b7dccd31
updates generate moshi asr
72429523
moshi asr integration tests (single + batched)
135730e7
update tests
0679d29a
update conversion script
99322b76
good default sliding window value
427bfb0b
udpdate generate
18204575
update test checkpoint
12674591
nit
78a5c677
fix mimi
08ee7aaa
fix codec prefix
29cd507b
Merge branch 'huggingface:main' into moshi-asr
34b9b931
revert
ac460b7a
revert
ba800828
update config
1b75b130
update config
b2c7b317
unnecessary mimi input restriction
75fa51e0
remove delay in tokens
2f9b96e0
remove _prepare_4d_causal_attention_mask_with_cache_position and _upd…
fa42f380
test update
f65960f8
modular update
1541d27b
make style
43255cda
resolve merge conflict
694b202b
nit
73386fcb
rename
8676cc4e
create codec model generation config at init
92773156
remove delay
45820e40
max_new_tokens/length warning
0d4e86ee
correct conv1 padding cache import for modular
e75bfaf2
nit
076761e4
fix on encoder_past_key_values
6bd979f1
convert modular
14694059
move frame_size to config
c7f8e350
move frame_size to config
d37596d1
ArthurZucker
approved these changes
on 2025-06-19
update test name
74af6ad6
handle first token is bos
d7820af2
better handling of max_new_tokens
0edf0a54
fix
496285d9
Merge branch 'main' into add-kyutai-stt
9e277cfa
fix batch size in test input prep
b3774f43
update docstring
45054525
convert modular
f6f5adbd
make style
3922c8d5
Merge branch 'main' into add-kyutai-stt
b31f3d95
make style
48581547
Merge branch 'main' into add-kyutai-stt
7dbac18e
add feature extractor
103a6e79
correct modular convention name for feature_extraction file
f50d3644
update convertion script
3b630574
doc processor
b0199e30
update doc
df39c44f
udpate init
17ca2352
update model type
3eeee3da
fixes
07385114
update tests
0cbfab9f
fix
1e016c8b
make
25e6ddef
Merge branch 'main' into add-kyutai-stt
dd429a42
add doc
01ceb4bb
nit
cadfd6fe
fix
aa11ab03
doc
ea542d28
auto mappings
023748e2
doc
0d327f0e
nit
d2a2802a
convert modular
f326b978
doc
67ae5c00
nit
8ee71534
extend _keep_in_fp32_modules to enforce fp32
f46bd171
renaming to stt
e36504a8
doc update + test update
1f7685f3
doc fixes
fb5f8e38
doc fix
83b2366c
Merge branch 'main' into add-kyutai-stt
48c7d7b0
doc fix
5b19257a
fix musicgen tests
0725a4b2
fix musicgen tests
06f5ebf6
make style
0ef59b8e
fix musicgen tests
b11a6351
correct frame_rate config param for mimi
a669b8e8
update mimi test
ad3c364d
revert update mimi test
3ae13c97
enforce cpu test
faa4396f
move cache init in cache class
002e7fe4
convert modular
a9a46b6b
docstring update
b52dd43b
Merge branch 'main' into add-kyutai-stt
2251e9d3
update model id
d6292fc6
feature_extractor -> feature_extraction (SEW)
63fe49cf
convert modular
3bf80737
Merge branch 'main' into add-kyutai-stt
be61fcdc
disabled auto-merge 185 days ago
Manually disabled by user
Manually disabled by user
update model id
c9ddfef7
disabled auto-merge 185 days ago
Manually disabled by user
Manually disabled by user
Login to write a write a comment.
Login via GitHub