Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -57,15 +57,55 @@ class calculateDuration:
|
|
57 |
|
58 |
@spaces.GPU(duration=120)
|
59 |
@torch.inference_mode()
|
60 |
-
def generate_image(prompt, adapter_names, steps, seed, cfg_scale, width, height, progress):
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
gr.Info("Start to generate images ...")
|
63 |
|
64 |
with calculateDuration(f"Make a new generator:{seed}"):
|
65 |
pipe.to(device)
|
66 |
generator = torch.Generator(device=device).manual_seed(seed)
|
|
|
67 |
if len(adapter_names) > 0:
|
68 |
pipe.fuse_lora(adapter_names=adapter_names)
|
|
|
69 |
with calculateDuration("Generating image"):
|
70 |
# Generate image
|
71 |
generated_image = pipe(
|
@@ -133,46 +173,12 @@ def run_lora(prompt, lora_strings_json, cfg_scale, steps, randomize_seed, seed,
|
|
133 |
|
134 |
# Load LoRA weights
|
135 |
gr.Info("Start to load loras ...")
|
136 |
-
|
137 |
-
lora_configs = None
|
138 |
-
adapter_names = []
|
139 |
-
if lora_strings_json:
|
140 |
-
try:
|
141 |
-
lora_configs = json.loads(lora_strings_json)
|
142 |
-
except:
|
143 |
-
gr.Warning("Parse lora config json failed")
|
144 |
-
print("parse lora config json failed")
|
145 |
-
|
146 |
-
if lora_configs:
|
147 |
-
with calculateDuration("Loading LoRA weights"):
|
148 |
-
|
149 |
-
adapter_weights = []
|
150 |
-
for lora_info in lora_configs:
|
151 |
-
lora_repo = lora_info.get("repo")
|
152 |
-
weights = lora_info.get("weights")
|
153 |
-
adapter_name = lora_info.get("adapter_name")
|
154 |
-
adapter_weight = lora_info.get("adapter_weight")
|
155 |
-
|
156 |
-
adapter_names.append(adapter_name)
|
157 |
-
adapter_weights.append(adapter_weight)
|
158 |
-
|
159 |
-
if lora_repo and weights and adapter_name:
|
160 |
-
# load lora
|
161 |
-
try:
|
162 |
-
pipe.load_lora_weights(lora_repo, weight_name=weights, adapter_name=adapter_name)
|
163 |
-
except ValueError as e:
|
164 |
-
print(f"Error loading LoRA adapter: {e}")
|
165 |
-
continue
|
166 |
-
|
167 |
-
# set lora weights
|
168 |
-
if len(adapter_names) > 0:
|
169 |
-
pipe.set_adapters(adapter_names, adapter_weights=adapter_weights)
|
170 |
-
|
171 |
# Generate image
|
172 |
error_message = ""
|
173 |
try:
|
174 |
print("Start applying for zeroGPU resources")
|
175 |
-
final_image = generate_image(prompt, adapter_names, steps, seed, cfg_scale, width, height, progress)
|
176 |
except Exception as e:
|
177 |
error_message = str(e)
|
178 |
gr.Error(error_message)
|
|
|
57 |
|
58 |
@spaces.GPU(duration=120)
|
59 |
@torch.inference_mode()
|
60 |
+
def generate_image(prompt, lora_strings_json, adapter_names, steps, seed, cfg_scale, width, height, progress):
|
61 |
|
62 |
+
lora_configs = None
|
63 |
+
adapter_names = []
|
64 |
+
if lora_strings_json:
|
65 |
+
try:
|
66 |
+
lora_configs = json.loads(lora_strings_json)
|
67 |
+
except:
|
68 |
+
gr.Warning("Parse lora config json failed")
|
69 |
+
print("parse lora config json failed")
|
70 |
+
|
71 |
+
if lora_configs:
|
72 |
+
with calculateDuration("Loading LoRA weights"):
|
73 |
+
adapter_weights = []
|
74 |
+
for lora_info in lora_configs:
|
75 |
+
lora_repo = lora_info.get("repo")
|
76 |
+
weights = lora_info.get("weights")
|
77 |
+
adapter_name = lora_info.get("adapter_name")
|
78 |
+
adapter_weight = lora_info.get("adapter_weight")
|
79 |
+
|
80 |
+
if lora_repo and weights and adapter_name:
|
81 |
+
retry_count = 3
|
82 |
+
for attempt in range(retry_count):
|
83 |
+
try:
|
84 |
+
pipe.load_lora_weights(lora_repo, weight_name=weights, adapter_name=adapter_name)
|
85 |
+
adapter_names.append(adapter_name)
|
86 |
+
adapter_weights.append(adapter_weight)
|
87 |
+
break # Load successful, exit retry loop
|
88 |
+
except ValueError as e:
|
89 |
+
print(f"Attempt {attempt+1}/{retry_count} failed to load LoRA adapter: {e}")
|
90 |
+
if attempt == retry_count - 1:
|
91 |
+
print(f"Error loading LoRA adapter: {adapter_name} after {retry_count} attempts")
|
92 |
+
else:
|
93 |
+
time.sleep(1) # Wait before retrying
|
94 |
+
|
95 |
+
# set lora weights
|
96 |
+
if len(adapter_names) > 0:
|
97 |
+
pipe.set_adapters(adapter_names, adapter_weights=adapter_weights)
|
98 |
+
|
99 |
+
|
100 |
gr.Info("Start to generate images ...")
|
101 |
|
102 |
with calculateDuration(f"Make a new generator:{seed}"):
|
103 |
pipe.to(device)
|
104 |
generator = torch.Generator(device=device).manual_seed(seed)
|
105 |
+
|
106 |
if len(adapter_names) > 0:
|
107 |
pipe.fuse_lora(adapter_names=adapter_names)
|
108 |
+
|
109 |
with calculateDuration("Generating image"):
|
110 |
# Generate image
|
111 |
generated_image = pipe(
|
|
|
173 |
|
174 |
# Load LoRA weights
|
175 |
gr.Info("Start to load loras ...")
|
176 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
177 |
# Generate image
|
178 |
error_message = ""
|
179 |
try:
|
180 |
print("Start applying for zeroGPU resources")
|
181 |
+
final_image = generate_image(prompt, lora_strings_json, adapter_names, steps, seed, cfg_scale, width, height, progress)
|
182 |
except Exception as e:
|
183 |
error_message = str(e)
|
184 |
gr.Error(error_message)
|