Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -44,7 +44,14 @@ class model_onxx:
|
|
44 |
def __init__(self):
|
45 |
self.model=None
|
46 |
self.n_onxx=""
|
|
|
47 |
pass
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
def function_change(self,n_model,token,n_onxx,choice):
|
49 |
if choice=="decoder":
|
50 |
|
@@ -66,12 +73,15 @@ class model_onxx:
|
|
66 |
def convert_to_onnx_only_decoder(self,n_model,token,namemodelonxx):
|
67 |
model=VitsModel.from_pretrained(n_model,token=token)
|
68 |
x=f"{namemodelonxx}.onnx"
|
|
|
|
|
|
|
69 |
vocab_size = model.text_encoder.embed_tokens.weight.size(0)
|
70 |
example_input = torch.randint(0, vocab_size, (1, 100), dtype=torch.long)
|
71 |
torch.onnx.export(
|
72 |
model, # The model to be exported
|
73 |
example_input, # Example input for the model
|
74 |
-
|
75 |
opset_version=11, # Use an appropriate ONNX opset version
|
76 |
input_names=['input'], # Name of the input layer
|
77 |
output_names=['output'], # Name of the output layer
|
@@ -80,7 +90,7 @@ class model_onxx:
|
|
80 |
'output': {0: 'batch_size'}
|
81 |
}
|
82 |
)
|
83 |
-
return
|
84 |
def convert_to_onnx_all(self,n_model,token ,namemodelonxx):
|
85 |
|
86 |
model=VitsModel.from_pretrained(n_model,token=token)
|
|
|
44 |
def __init__(self):
|
45 |
self.model=None
|
46 |
self.n_onxx=""
|
47 |
+
self.storage_dir = "uploads"
|
48 |
pass
|
49 |
+
|
50 |
+
|
51 |
+
|
52 |
+
|
53 |
+
|
54 |
+
|
55 |
def function_change(self,n_model,token,n_onxx,choice):
|
56 |
if choice=="decoder":
|
57 |
|
|
|
73 |
def convert_to_onnx_only_decoder(self,n_model,token,namemodelonxx):
|
74 |
model=VitsModel.from_pretrained(n_model,token=token)
|
75 |
x=f"{namemodelonxx}.onnx"
|
76 |
+
if not os.path.exists("uploads"):
|
77 |
+
os.makedirs(storage_dir)
|
78 |
+
file_path = os.path.join("uploads",x)
|
79 |
vocab_size = model.text_encoder.embed_tokens.weight.size(0)
|
80 |
example_input = torch.randint(0, vocab_size, (1, 100), dtype=torch.long)
|
81 |
torch.onnx.export(
|
82 |
model, # The model to be exported
|
83 |
example_input, # Example input for the model
|
84 |
+
file_path, # The filename for the exported ONNX model
|
85 |
opset_version=11, # Use an appropriate ONNX opset version
|
86 |
input_names=['input'], # Name of the input layer
|
87 |
output_names=['output'], # Name of the output layer
|
|
|
90 |
'output': {0: 'batch_size'}
|
91 |
}
|
92 |
)
|
93 |
+
return file_path
|
94 |
def convert_to_onnx_all(self,n_model,token ,namemodelonxx):
|
95 |
|
96 |
model=VitsModel.from_pretrained(n_model,token=token)
|