Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -51,9 +51,7 @@ print(f"Reading {demo_script_path} to apply environment-specific modifications..
|
|
51 |
try:
|
52 |
modified_content = demo_script_path.read_text()
|
53 |
|
54 |
-
#
|
55 |
-
|
56 |
-
# Define the original model loading block to be replaced.
|
57 |
original_model_lines = [
|
58 |
' self.model = VibeVoiceForConditionalGenerationInference.from_pretrained(',
|
59 |
' self.model_path,',
|
@@ -63,19 +61,10 @@ try:
|
|
63 |
' )'
|
64 |
]
|
65 |
original_model_block = "\n".join(original_model_lines)
|
66 |
-
|
67 |
-
#
|
68 |
-
|
69 |
-
|
70 |
-
' num_speakers: int,',
|
71 |
-
' script: str,',
|
72 |
-
' speaker_1: str = None,',
|
73 |
-
' speaker_2: str = None,',
|
74 |
-
' speaker_3: str = None,',
|
75 |
-
' speaker_4: str = None,',
|
76 |
-
' cfg_scale: float = 1.3) -> Iterator[tuple]:'
|
77 |
-
]
|
78 |
-
original_method_signature = "\n".join(original_method_lines)
|
79 |
|
80 |
if USE_ZEROGPU:
|
81 |
print("Optimizing for ZeroGPU execution...")
|
@@ -94,8 +83,8 @@ try:
|
|
94 |
]
|
95 |
replacement_model_block_gpu = "\n".join(replacement_model_lines_gpu)
|
96 |
|
97 |
-
# Add the @spaces.GPU decorator
|
98 |
-
replacement_method_signature_gpu = "@spaces.GPU(duration=120)\n" + original_method_signature
|
99 |
|
100 |
# --- Apply Patches for GPU ---
|
101 |
|
@@ -104,14 +93,16 @@ try:
|
|
104 |
modified_content = modified_content.replace(original_method_signature, replacement_method_signature_gpu)
|
105 |
print("Successfully applied GPU decorator to the generation method.")
|
106 |
else:
|
107 |
-
print("\033[
|
|
|
108 |
|
109 |
# Patch 2: Modify the model loading
|
110 |
if original_model_block in modified_content:
|
111 |
modified_content = modified_content.replace(original_model_block, replacement_model_block_gpu)
|
112 |
print("Successfully patched the model loading block for ZeroGPU.")
|
113 |
else:
|
114 |
-
print("\033[
|
|
|
115 |
|
116 |
else: # Pure CPU execution
|
117 |
print("Modifying for pure CPU execution...")
|
@@ -131,8 +122,8 @@ try:
|
|
131 |
modified_content = modified_content.replace(original_model_block, replacement_model_block_cpu)
|
132 |
print("Script modified for CPU successfully.")
|
133 |
else:
|
134 |
-
print("\033[
|
135 |
-
|
136 |
|
137 |
# Write the dynamically modified content back to the demo file
|
138 |
demo_script_path.write_text(modified_content)
|
@@ -142,7 +133,7 @@ except Exception as e:
|
|
142 |
sys.exit(1)
|
143 |
|
144 |
# --- 4. Launch the Gradio Demo ---
|
145 |
-
model_id = "microsoft/
|
146 |
|
147 |
# Construct the command to run the modified demo script
|
148 |
command = [
|
|
|
51 |
try:
|
52 |
modified_content = demo_script_path.read_text()
|
53 |
|
54 |
+
# Define the original model loading block using a list of lines for robustness.
|
|
|
|
|
55 |
original_model_lines = [
|
56 |
' self.model = VibeVoiceForConditionalGenerationInference.from_pretrained(',
|
57 |
' self.model_path,',
|
|
|
61 |
' )'
|
62 |
]
|
63 |
original_model_block = "\n".join(original_model_lines)
|
64 |
+
|
65 |
+
# More robustly define the generation method signature to patch.
|
66 |
+
# We only need the first line to find our target.
|
67 |
+
original_method_signature = " def generate_podcast_streaming(self,"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
|
69 |
if USE_ZEROGPU:
|
70 |
print("Optimizing for ZeroGPU execution...")
|
|
|
83 |
]
|
84 |
replacement_model_block_gpu = "\n".join(replacement_model_lines_gpu)
|
85 |
|
86 |
+
# Add the @spaces.GPU decorator *with correct indentation* before the method.
|
87 |
+
replacement_method_signature_gpu = " @spaces.GPU(duration=120)\n" + original_method_signature
|
88 |
|
89 |
# --- Apply Patches for GPU ---
|
90 |
|
|
|
93 |
modified_content = modified_content.replace(original_method_signature, replacement_method_signature_gpu)
|
94 |
print("Successfully applied GPU decorator to the generation method.")
|
95 |
else:
|
96 |
+
print("\033[91mError: Could not find the generation method signature to apply the GPU decorator.\033[0m")
|
97 |
+
sys.exit(1)
|
98 |
|
99 |
# Patch 2: Modify the model loading
|
100 |
if original_model_block in modified_content:
|
101 |
modified_content = modified_content.replace(original_model_block, replacement_model_block_gpu)
|
102 |
print("Successfully patched the model loading block for ZeroGPU.")
|
103 |
else:
|
104 |
+
print("\033[91mError: The original model loading block was not found. Patching may have failed.\033[0m")
|
105 |
+
sys.exit(1)
|
106 |
|
107 |
else: # Pure CPU execution
|
108 |
print("Modifying for pure CPU execution...")
|
|
|
122 |
modified_content = modified_content.replace(original_model_block, replacement_model_block_cpu)
|
123 |
print("Script modified for CPU successfully.")
|
124 |
else:
|
125 |
+
print("\033[91mError: The original model loading block was not found. Patching may have failed.\03-3[0m")
|
126 |
+
sys.exit(1)
|
127 |
|
128 |
# Write the dynamically modified content back to the demo file
|
129 |
demo_script_path.write_text(modified_content)
|
|
|
133 |
sys.exit(1)
|
134 |
|
135 |
# --- 4. Launch the Gradio Demo ---
|
136 |
+
model_id = "microsoft/VibeVoice-1.5B"
|
137 |
|
138 |
# Construct the command to run the modified demo script
|
139 |
command = [
|