Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -10,7 +10,6 @@ import json
|
|
10 |
import logging
|
11 |
from dataclasses import dataclass
|
12 |
import gc
|
13 |
-
import os
|
14 |
|
15 |
# Configure logging
|
16 |
logging.basicConfig(
|
@@ -30,10 +29,6 @@ class ModelCache:
|
|
30 |
def __init__(self, cache_dir: Path):
|
31 |
self.cache_dir = cache_dir
|
32 |
self.cache_dir.mkdir(parents=True, exist_ok=True)
|
33 |
-
|
34 |
-
# Set environment variables for better memory management
|
35 |
-
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:512'
|
36 |
-
os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
|
37 |
|
38 |
def load_model(self, model_id: str, load_func: callable, cache_name: str) -> Any:
|
39 |
try:
|
@@ -52,32 +47,18 @@ class EnhancedBanglaSDGenerator:
|
|
52 |
):
|
53 |
self.device = device or torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
54 |
logger.info(f"Using device: {self.device}")
|
55 |
-
|
56 |
-
# Set memory split for VRAM usage on CPU
|
57 |
-
self.memory_split = 0.5 # Use 50% of available VRAM
|
58 |
-
self.setup_memory_management()
|
59 |
-
|
60 |
self.cache = ModelCache(Path(cache_dir))
|
61 |
self._initialize_models(banglaclip_weights_path)
|
62 |
self._load_context_data()
|
63 |
|
64 |
-
def setup_memory_management(self):
|
65 |
-
"""Setup optimal memory management for CPU and VRAM"""
|
66 |
-
if torch.cuda.is_available():
|
67 |
-
total_memory = torch.cuda.get_device_properties(0).total_memory
|
68 |
-
torch.cuda.set_per_process_memory_fraction(self.memory_split)
|
69 |
-
|
70 |
-
# Optimize CPU memory
|
71 |
-
torch.set_num_threads(min(8, os.cpu_count() or 4))
|
72 |
-
torch.set_num_interop_threads(min(8, os.cpu_count() or 4))
|
73 |
-
|
74 |
def _initialize_models(self, banglaclip_weights_path: str):
|
75 |
try:
|
76 |
# Initialize translation models
|
77 |
self.bn2en_model_name = "Helsinki-NLP/opus-mt-bn-en"
|
78 |
self.translator = self.cache.load_model(
|
79 |
self.bn2en_model_name,
|
80 |
-
|
81 |
"translator"
|
82 |
).to(self.device)
|
83 |
self.trans_tokenizer = MarianTokenizer.from_pretrained(self.bn2en_model_name)
|
@@ -98,52 +79,34 @@ class EnhancedBanglaSDGenerator:
|
|
98 |
|
99 |
def _initialize_stable_diffusion(self):
|
100 |
"""Initialize Stable Diffusion pipeline with optimized settings."""
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
),
|
111 |
-
"stable_diffusion"
|
112 |
-
)
|
113 |
|
114 |
-
|
115 |
-
self.pipe.scheduler
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
)
|
121 |
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
self.pipe.enable_sequential_cpu_offload()
|
126 |
-
|
127 |
-
# VRAM optimization
|
128 |
-
if torch.cuda.is_available():
|
129 |
-
torch.cuda.empty_cache()
|
130 |
-
self.pipe.enable_model_cpu_offload()
|
131 |
-
|
132 |
-
self.pipe = self.pipe.to(self.device)
|
133 |
-
|
134 |
-
except Exception as e:
|
135 |
-
logger.error(f"Error initializing Stable Diffusion: {str(e)}")
|
136 |
-
raise
|
137 |
|
138 |
def _load_banglaclip_model(self, weights_path: str) -> CLIPModel:
|
139 |
try:
|
140 |
if not Path(weights_path).exists():
|
141 |
raise FileNotFoundError(f"BanglaCLIP weights not found at {weights_path}")
|
142 |
|
143 |
-
clip_model = CLIPModel.from_pretrained(
|
144 |
-
self.clip_model_name,
|
145 |
-
low_cpu_mem_usage=True
|
146 |
-
)
|
147 |
state_dict = torch.load(weights_path, map_location=self.device)
|
148 |
|
149 |
cleaned_state_dict = {
|
@@ -178,12 +141,22 @@ class EnhancedBanglaSDGenerator:
|
|
178 |
inputs = self.trans_tokenizer(bangla_text, return_tensors="pt", padding=True)
|
179 |
inputs = {k: v.to(self.device) for k, v in inputs.items()}
|
180 |
|
181 |
-
with torch.no_grad()
|
182 |
outputs = self.translator.generate(**inputs)
|
183 |
|
184 |
translated = self.trans_tokenizer.decode(outputs[0], skip_special_tokens=True)
|
185 |
return translated
|
186 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
187 |
def generate_image(
|
188 |
self,
|
189 |
bangla_text: str,
|
@@ -198,29 +171,18 @@ class EnhancedBanglaSDGenerator:
|
|
198 |
if config.seed is not None:
|
199 |
torch.manual_seed(config.seed)
|
200 |
|
201 |
-
# Clear memory before generation
|
202 |
-
gc.collect()
|
203 |
-
torch.cuda.empty_cache() if torch.cuda.is_available() else None
|
204 |
-
|
205 |
enhanced_prompt = self._enhance_prompt(bangla_text)
|
206 |
negative_prompt = self._get_negative_prompt()
|
207 |
|
208 |
-
|
209 |
-
with torch.inference_mode(), torch.cpu.amp.autocast():
|
210 |
result = self.pipe(
|
211 |
prompt=enhanced_prompt,
|
212 |
negative_prompt=negative_prompt,
|
213 |
num_images_per_prompt=config.num_images,
|
214 |
num_inference_steps=config.num_inference_steps,
|
215 |
-
guidance_scale=config.guidance_scale
|
216 |
-
use_memory_efficient_attention=True,
|
217 |
-
use_memory_efficient_cross_attention=True
|
218 |
)
|
219 |
|
220 |
-
# Clear memory after generation
|
221 |
-
gc.collect()
|
222 |
-
torch.cuda.empty_cache() if torch.cuda.is_available() else None
|
223 |
-
|
224 |
return result.images, enhanced_prompt
|
225 |
|
226 |
except Exception as e:
|
@@ -231,10 +193,12 @@ class EnhancedBanglaSDGenerator:
|
|
231 |
"""Enhance prompt with context and style information."""
|
232 |
translated_text = self._translate_text(bangla_text)
|
233 |
|
|
|
234 |
contexts = []
|
235 |
contexts.extend(context for loc, context in self.location_contexts.items() if loc in bangla_text)
|
236 |
contexts.extend(context for scene, context in self.scene_contexts.items() if scene in bangla_text)
|
237 |
|
|
|
238 |
photo_style = [
|
239 |
"professional photography",
|
240 |
"high resolution",
|
@@ -244,6 +208,7 @@ class EnhancedBanglaSDGenerator:
|
|
244 |
"beautiful composition"
|
245 |
]
|
246 |
|
|
|
247 |
all_parts = [translated_text] + contexts + photo_style
|
248 |
return ", ".join(dict.fromkeys(all_parts))
|
249 |
|
@@ -352,9 +317,6 @@ def create_gradio_interface():
|
|
352 |
return demo
|
353 |
|
354 |
if __name__ == "__main__":
|
355 |
-
# Set environment variables for better performance
|
356 |
-
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:512'
|
357 |
-
os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
|
358 |
-
|
359 |
demo = create_gradio_interface()
|
|
|
360 |
demo.queue().launch(share=True)
|
|
|
10 |
import logging
|
11 |
from dataclasses import dataclass
|
12 |
import gc
|
|
|
13 |
|
14 |
# Configure logging
|
15 |
logging.basicConfig(
|
|
|
29 |
def __init__(self, cache_dir: Path):
|
30 |
self.cache_dir = cache_dir
|
31 |
self.cache_dir.mkdir(parents=True, exist_ok=True)
|
|
|
|
|
|
|
|
|
32 |
|
33 |
def load_model(self, model_id: str, load_func: callable, cache_name: str) -> Any:
|
34 |
try:
|
|
|
47 |
):
|
48 |
self.device = device or torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
49 |
logger.info(f"Using device: {self.device}")
|
50 |
+
|
|
|
|
|
|
|
|
|
51 |
self.cache = ModelCache(Path(cache_dir))
|
52 |
self._initialize_models(banglaclip_weights_path)
|
53 |
self._load_context_data()
|
54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
def _initialize_models(self, banglaclip_weights_path: str):
|
56 |
try:
|
57 |
# Initialize translation models
|
58 |
self.bn2en_model_name = "Helsinki-NLP/opus-mt-bn-en"
|
59 |
self.translator = self.cache.load_model(
|
60 |
self.bn2en_model_name,
|
61 |
+
MarianMTModel.from_pretrained,
|
62 |
"translator"
|
63 |
).to(self.device)
|
64 |
self.trans_tokenizer = MarianTokenizer.from_pretrained(self.bn2en_model_name)
|
|
|
79 |
|
80 |
def _initialize_stable_diffusion(self):
|
81 |
"""Initialize Stable Diffusion pipeline with optimized settings."""
|
82 |
+
self.pipe = self.cache.load_model(
|
83 |
+
"runwayml/stable-diffusion-v1-5",
|
84 |
+
lambda model_id: StableDiffusionPipeline.from_pretrained(
|
85 |
+
model_id,
|
86 |
+
torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
|
87 |
+
safety_checker=None
|
88 |
+
),
|
89 |
+
"stable_diffusion"
|
90 |
+
)
|
|
|
|
|
|
|
91 |
|
92 |
+
self.pipe.scheduler = DPMSolverMultistepScheduler.from_config(
|
93 |
+
self.pipe.scheduler.config,
|
94 |
+
use_karras_sigmas=True,
|
95 |
+
algorithm_type="dpmsolver++"
|
96 |
+
)
|
97 |
+
self.pipe = self.pipe.to(self.device)
|
|
|
98 |
|
99 |
+
# Memory optimization
|
100 |
+
self.pipe.enable_attention_slicing()
|
101 |
+
if torch.cuda.is_available():
|
102 |
self.pipe.enable_sequential_cpu_offload()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
103 |
|
104 |
def _load_banglaclip_model(self, weights_path: str) -> CLIPModel:
|
105 |
try:
|
106 |
if not Path(weights_path).exists():
|
107 |
raise FileNotFoundError(f"BanglaCLIP weights not found at {weights_path}")
|
108 |
|
109 |
+
clip_model = CLIPModel.from_pretrained(self.clip_model_name)
|
|
|
|
|
|
|
110 |
state_dict = torch.load(weights_path, map_location=self.device)
|
111 |
|
112 |
cleaned_state_dict = {
|
|
|
141 |
inputs = self.trans_tokenizer(bangla_text, return_tensors="pt", padding=True)
|
142 |
inputs = {k: v.to(self.device) for k, v in inputs.items()}
|
143 |
|
144 |
+
with torch.no_grad():
|
145 |
outputs = self.translator.generate(**inputs)
|
146 |
|
147 |
translated = self.trans_tokenizer.decode(outputs[0], skip_special_tokens=True)
|
148 |
return translated
|
149 |
|
150 |
+
def _get_text_embedding(self, text: str):
|
151 |
+
"""Get text embedding from BanglaCLIP model."""
|
152 |
+
inputs = self.tokenizer(text, return_tensors="pt", padding=True, truncation=True)
|
153 |
+
inputs = {k: v.to(self.device) for k, v in inputs.items()}
|
154 |
+
|
155 |
+
with torch.no_grad():
|
156 |
+
outputs = self.banglaclip_model.get_text_features(**inputs)
|
157 |
+
|
158 |
+
return outputs
|
159 |
+
|
160 |
def generate_image(
|
161 |
self,
|
162 |
bangla_text: str,
|
|
|
171 |
if config.seed is not None:
|
172 |
torch.manual_seed(config.seed)
|
173 |
|
|
|
|
|
|
|
|
|
174 |
enhanced_prompt = self._enhance_prompt(bangla_text)
|
175 |
negative_prompt = self._get_negative_prompt()
|
176 |
|
177 |
+
with torch.autocast(self.device.type):
|
|
|
178 |
result = self.pipe(
|
179 |
prompt=enhanced_prompt,
|
180 |
negative_prompt=negative_prompt,
|
181 |
num_images_per_prompt=config.num_images,
|
182 |
num_inference_steps=config.num_inference_steps,
|
183 |
+
guidance_scale=config.guidance_scale
|
|
|
|
|
184 |
)
|
185 |
|
|
|
|
|
|
|
|
|
186 |
return result.images, enhanced_prompt
|
187 |
|
188 |
except Exception as e:
|
|
|
193 |
"""Enhance prompt with context and style information."""
|
194 |
translated_text = self._translate_text(bangla_text)
|
195 |
|
196 |
+
# Gather contexts
|
197 |
contexts = []
|
198 |
contexts.extend(context for loc, context in self.location_contexts.items() if loc in bangla_text)
|
199 |
contexts.extend(context for scene, context in self.scene_contexts.items() if scene in bangla_text)
|
200 |
|
201 |
+
# Add photo style
|
202 |
photo_style = [
|
203 |
"professional photography",
|
204 |
"high resolution",
|
|
|
208 |
"beautiful composition"
|
209 |
]
|
210 |
|
211 |
+
# Combine all parts
|
212 |
all_parts = [translated_text] + contexts + photo_style
|
213 |
return ", ".join(dict.fromkeys(all_parts))
|
214 |
|
|
|
317 |
return demo
|
318 |
|
319 |
if __name__ == "__main__":
|
|
|
|
|
|
|
|
|
320 |
demo = create_gradio_interface()
|
321 |
+
# Fixed queue configuration for newer Gradio versions
|
322 |
demo.queue().launch(share=True)
|