Add SeamlessM4T v2 #27779
add working convertion script
2c220bf6
first non-working version of modeling code
7ae2c535
update modeling code (working)
4fab4037
make style
1d48949d
make fix-copies
a51bd67d
add config docstrings
4472212a
add config to ignore docstrings formatage due to unconventional markdown
a07b5d86
fix copies
1cdcd1ff
fix generation num_return_sequences
00a47080
enrich docs
181214cc
add and fix tests beside integration tests
5dd8a32b
update integration tests
fa37cdc8
update repo id
4799132d
add tie weights and make style
b52046b8
correct naming in .md
b0b2fe50
fix imports and so on
47b7a3ed
correct docstrings
9172254d
fix fp16 speech forward
02147ab5
fix speechencoder attention
fc6dd5d8
make style
bebc1459
fix copied from
303b5487
rename SeamlessM4Tv2-v2 to SeamlessM4Tv2
08410ece
Apply suggestions on configuration
7f6fbbac
remove useless public models
92a7600f
fix private models + better naming for T2U models
c8b83eb4
clean speech encoder relative position embeddings
37ad443a
refactor chunk attention
02b8c763
add docstrings to chunk attention method
e510f590
improve naming and docstrings
bbbcbeb3
rename some attention variables + add temperature sampling in T2U model
f81b74e1
rename DOCSTRINGS variable names
adee1dbf
make style + remove 2 useless config parameters
1e8f70e2
enrich model card
743b1609
remove any attention_head reference + fix temperature in T2U
f88e9e02
new fmt and make style
110e82ed
Apply suggestions from code review
e36da6c4
rename spkr_id->speaker_id and change docstrings of get_char_input_ids
d89101db
simplify v2attention
bec648b0
make style
d73a7fef
Update seamless_m4t_v2.md
7fadf0ef
update code and tests with last update
e9cbe8a6
update repo ids
ce326a52
Merge branch 'huggingface:main' into add-S2S-2
2ff8fbe9
fill article name, abstract andauthors
fb8a66f4
update not_doctested and slow_doc tests
46a86c6a
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub