Spaces:
Runtime error
Runtime error
aleegr10
commited on
Commit
·
5fa1d66
1
Parent(s):
3d39e44
update
Browse files- app.py +2 -3
- requirements.txt +10 -2
app.py
CHANGED
|
@@ -49,7 +49,7 @@ def modelo3En(text):
|
|
| 49 |
|
| 50 |
speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
|
| 51 |
|
| 52 |
-
return gr.Audio
|
| 53 |
|
| 54 |
|
| 55 |
def modelo3Es(text):
|
|
@@ -64,11 +64,10 @@ def modelo3Es(text):
|
|
| 64 |
|
| 65 |
speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
|
| 66 |
|
| 67 |
-
return gr.Audio
|
| 68 |
|
| 69 |
|
| 70 |
def executionIMG(image, lan):
|
| 71 |
-
print(lan)
|
| 72 |
if lan == 'english':
|
| 73 |
model1res = modelo1(image)
|
| 74 |
model3res = modelo3En(model1res)
|
|
|
|
| 49 |
|
| 50 |
speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
|
| 51 |
|
| 52 |
+
return gr.Audio(value=(16000, speech.cpu().numpy()))
|
| 53 |
|
| 54 |
|
| 55 |
def modelo3Es(text):
|
|
|
|
| 64 |
|
| 65 |
speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
|
| 66 |
|
| 67 |
+
return gr.Audio(value=(16000, speech.cpu().numpy()))
|
| 68 |
|
| 69 |
|
| 70 |
def executionIMG(image, lan):
|
|
|
|
| 71 |
if lan == 'english':
|
| 72 |
model1res = modelo1(image)
|
| 73 |
model3res = modelo3En(model1res)
|
requirements.txt
CHANGED
|
@@ -11,6 +11,7 @@ certifi==2023.7.22
|
|
| 11 |
cffi==1.16.0
|
| 12 |
charset-normalizer==3.3.1
|
| 13 |
click==8.1.7
|
|
|
|
| 14 |
contourpy==1.1.1
|
| 15 |
cycler==0.12.1
|
| 16 |
datasets==2.14.6
|
|
@@ -23,8 +24,8 @@ filelock==3.13.0
|
|
| 23 |
fonttools==4.43.1
|
| 24 |
frozenlist==1.4.0
|
| 25 |
fsspec==2023.10.0
|
| 26 |
-
gradio==
|
| 27 |
-
gradio_client==0.
|
| 28 |
h11==0.14.0
|
| 29 |
httpcore==0.18.0
|
| 30 |
httpx==0.25.0
|
|
@@ -39,8 +40,10 @@ kiwisolver==1.4.5
|
|
| 39 |
lazy_loader==0.3
|
| 40 |
librosa==0.10.1
|
| 41 |
llvmlite==0.41.1
|
|
|
|
| 42 |
MarkupSafe==2.1.3
|
| 43 |
matplotlib==3.8.0
|
|
|
|
| 44 |
mpmath==1.3.0
|
| 45 |
msgpack==1.0.7
|
| 46 |
multidict==6.0.4
|
|
@@ -72,6 +75,7 @@ pycparser==2.21
|
|
| 72 |
pydantic==2.4.2
|
| 73 |
pydantic_core==2.10.1
|
| 74 |
pydub==0.25.1
|
|
|
|
| 75 |
pyparsing==3.1.1
|
| 76 |
python-dateutil==2.8.2
|
| 77 |
python-multipart==0.0.6
|
|
@@ -80,12 +84,14 @@ PyYAML==6.0.1
|
|
| 80 |
referencing==0.30.2
|
| 81 |
regex==2023.10.3
|
| 82 |
requests==2.31.0
|
|
|
|
| 83 |
rpds-py==0.10.6
|
| 84 |
safetensors==0.4.0
|
| 85 |
scikit-learn==1.3.2
|
| 86 |
scipy==1.11.3
|
| 87 |
semantic-version==2.10.0
|
| 88 |
sentencepiece==0.1.99
|
|
|
|
| 89 |
six==1.16.0
|
| 90 |
sniffio==1.3.0
|
| 91 |
soundfile==0.12.1
|
|
@@ -94,11 +100,13 @@ starlette==0.27.0
|
|
| 94 |
sympy==1.12
|
| 95 |
threadpoolctl==3.2.0
|
| 96 |
tokenizers==0.14.1
|
|
|
|
| 97 |
toolz==0.12.0
|
| 98 |
torch==2.1.0
|
| 99 |
tqdm==4.66.1
|
| 100 |
transformers==4.34.1
|
| 101 |
triton==2.1.0
|
|
|
|
| 102 |
typing_extensions==4.8.0
|
| 103 |
tzdata==2023.3
|
| 104 |
urllib3==2.0.7
|
|
|
|
| 11 |
cffi==1.16.0
|
| 12 |
charset-normalizer==3.3.1
|
| 13 |
click==8.1.7
|
| 14 |
+
colorama==0.4.6
|
| 15 |
contourpy==1.1.1
|
| 16 |
cycler==0.12.1
|
| 17 |
datasets==2.14.6
|
|
|
|
| 24 |
fonttools==4.43.1
|
| 25 |
frozenlist==1.4.0
|
| 26 |
fsspec==2023.10.0
|
| 27 |
+
gradio==4.1.1
|
| 28 |
+
gradio_client==0.7.0
|
| 29 |
h11==0.14.0
|
| 30 |
httpcore==0.18.0
|
| 31 |
httpx==0.25.0
|
|
|
|
| 40 |
lazy_loader==0.3
|
| 41 |
librosa==0.10.1
|
| 42 |
llvmlite==0.41.1
|
| 43 |
+
markdown-it-py==3.0.0
|
| 44 |
MarkupSafe==2.1.3
|
| 45 |
matplotlib==3.8.0
|
| 46 |
+
mdurl==0.1.2
|
| 47 |
mpmath==1.3.0
|
| 48 |
msgpack==1.0.7
|
| 49 |
multidict==6.0.4
|
|
|
|
| 75 |
pydantic==2.4.2
|
| 76 |
pydantic_core==2.10.1
|
| 77 |
pydub==0.25.1
|
| 78 |
+
Pygments==2.16.1
|
| 79 |
pyparsing==3.1.1
|
| 80 |
python-dateutil==2.8.2
|
| 81 |
python-multipart==0.0.6
|
|
|
|
| 84 |
referencing==0.30.2
|
| 85 |
regex==2023.10.3
|
| 86 |
requests==2.31.0
|
| 87 |
+
rich==13.6.0
|
| 88 |
rpds-py==0.10.6
|
| 89 |
safetensors==0.4.0
|
| 90 |
scikit-learn==1.3.2
|
| 91 |
scipy==1.11.3
|
| 92 |
semantic-version==2.10.0
|
| 93 |
sentencepiece==0.1.99
|
| 94 |
+
shellingham==1.5.4
|
| 95 |
six==1.16.0
|
| 96 |
sniffio==1.3.0
|
| 97 |
soundfile==0.12.1
|
|
|
|
| 100 |
sympy==1.12
|
| 101 |
threadpoolctl==3.2.0
|
| 102 |
tokenizers==0.14.1
|
| 103 |
+
tomlkit==0.12.0
|
| 104 |
toolz==0.12.0
|
| 105 |
torch==2.1.0
|
| 106 |
tqdm==4.66.1
|
| 107 |
transformers==4.34.1
|
| 108 |
triton==2.1.0
|
| 109 |
+
typer==0.9.0
|
| 110 |
typing_extensions==4.8.0
|
| 111 |
tzdata==2023.3
|
| 112 |
urllib3==2.0.7
|