refactoring

app.py

@@ -24,32 +24,31 @@ import pandas as pd
 import base64
 from config import *
 
-# code output order
-code = {}
-code_pos_device = '001_code'
-code_pos_data_type = '002_data_type'
-code_pos_tf32 = '003_tf32'
-code_pos_variant = '004_variant'
-code_pos_init_pipeline = '050_init_pipe'
-code_pos_requires_safety_checker = '054_requires_safety_checker'
-code_pos_safety_checker = '055_safety_checker'
-code_pos_scheduler = '060_scheduler'
-code_pos_generator = '070_generator'
-code_pos_prompt = '080_prompt'
-code_pos_negative_prompt = '085_negative_prompt'
-code_pos_inference_steps = '090_inference_steps'
-code_pos_guidance_scale = '095_guidance_scale'
-code_pos_run_inference = '100_run_inference'
-
-initial_config, devices, models, schedulers = get_inital_config()
+# get 
+# - initial configuration, 
+# - a list of available devices from the config file
+# - a list of available models from the config file
+# - a list of available schedulers from the config file
+# - a dict that contains code to for reproduction
+initial_config, devices, models, schedulers, code = get_inital_config()
+
+device                  = initial_config["device"]
+model                   = initial_config["model"]
+scheduler               = initial_config["scheduler"]
+variant                 = initial_config["variant"]
+allow_tensorfloat32     = initial_config["allow_tensorfloat32"]
+use_safetensors         = initial_config["use_safetensors"]
+data_type               = initial_config["data_type"]
+safety_checker          = initial_config["safety_checker"]
+requires_safety_checker = initial_config["requires_safety_checker"]
+manual_seed             = initial_config["manual_seed"]
+inference_steps         = initial_config["inference_steps"]
+guidance_scale          = initial_config["guidance_scale"]
+prompt                  = initial_config["prompt"]
+negative_prompt         = initial_config["negative_prompt"]
 
 config_history = []
 
-def get_sorted_code():
-    
-    return '\r\n'.join(value[1] for value in sorted(code.items()))
-
-# change methods
 def device_change(device):
     
     code[code_pos_device] = f'''device = "{device}"'''
@@ -225,26 +224,6 @@ def run_inference(model,
     
         return "Please select a model AND a scheduler.", None
 
-code[code_pos_device] = f'device = "{device}"'
-code[code_pos_variant] = f'variant = {variant}'
-code[code_pos_tf32] = f'torch.backends.cuda.matmul.allow_tf32 = {allow_tensorfloat32}'
-code[code_pos_data_type] = 'data_type = torch.bfloat16'
-code[code_pos_init_pipeline] = 'sys.exit("No model selected!")'
-code[code_pos_safety_checker] = 'pipeline.safety_checker = None'
-code[code_pos_requires_safety_checker] = f'pipeline.requires_safety_checker = {requires_safety_checker}'
-code[code_pos_scheduler] = 'sys.exit("No scheduler selected!")'
-code[code_pos_generator] = f'generator = torch.Generator("{device}")'
-code[code_pos_prompt] = f'prompt = "{prompt}"'
-code[code_pos_negative_prompt] = f'negative_prompt = "{negative_prompt}"'
-code[code_pos_inference_steps] = f'inference_steps = {inference_steps}'
-code[code_pos_guidance_scale] = f'guidance_scale = {guidance_scale}'
-code[code_pos_run_inference] = f'''image = pipeline(
-    prompt=prompt,
-    negative_prompt=negative_prompt,
-    generator=generator.manual_seed(manual_seed),
-    num_inference_steps=inference_steps, 
-    guidance_scale=guidance_scale).images[0]'''
-
 def dict_list_to_markdown_table(config_history):
 
     if not config_history:

config.py

@@ -3,6 +3,21 @@ import base64
 import json
 import torch
 
+code_pos_device = '001_code'
+code_pos_data_type = '002_data_type'
+code_pos_tf32 = '003_tf32'
+code_pos_variant = '004_variant'
+code_pos_init_pipeline = '050_init_pipe'
+code_pos_requires_safety_checker = '054_requires_safety_checker'
+code_pos_safety_checker = '055_safety_checker'
+code_pos_scheduler = '060_scheduler'
+code_pos_generator = '070_generator'
+code_pos_prompt = '080_prompt'
+code_pos_negative_prompt = '085_negative_prompt'
+code_pos_inference_steps = '090_inference_steps'
+code_pos_guidance_scale = '095_guidance_scale'
+code_pos_run_inference = '100_run_inference'
+
 def load_app_config():
     try:
         with open('appConfig.json', 'r') as f:
@@ -16,34 +31,26 @@ def load_app_config():
         
     return appConfig
 
-appConfig = load_app_config()
-
 def get_inital_config():
-        
-    # model config
+
+    appConfig = load_app_config()
+
+    # default model is None
     model_configs = appConfig.get("models", {})
+    # default model is None
     models = list(model_configs.keys())
     model = None
+    
+    # default scheduler is None
     scheduler_configs = appConfig.get("schedulers", {})
     schedulers = list(scheduler_configs.keys())
     scheduler = None
-
+    
+    # default device
     devices =  appConfig.get("devices", [])
     device = None
-
-    variant = None
-    allow_tensorfloat32 = False
-    use_safetensors = False
     data_type = 'float16'
-    safety_checker = False
-    requires_safety_checker = False
-    manual_seed = 42
-    inference_steps = 10
-    guidance_scale = 0.5
-    prompt = 'A white rabbit'
-    negative_prompt = 'lowres, cropped, worst quality, low quality, chat bubble, chat bubbles, ugly'
-
-    # # init device parameters
+    allow_tensorfloat32 = False
     if torch.cuda.is_available():
         device = "cuda" 
         data_type = 'bfloat16'
@@ -53,26 +60,47 @@ def get_inital_config():
     else:
         device = "cpu"
 
-    # inference config
     initial_config = {
         "device": device,
-        "model": model,
-        "scheduler": scheduler,
-        "variant": variant,
-        "allow_tensorflow": allow_tensorfloat32,
-        "use_safetensors": use_safetensors,
+        "model": None,
+        "scheduler": None,
+        "variant": None,
+        "allow_tensorflow32": allow_tensorfloat32,
+        "use_safetensors": False,
         "data_type": data_type,
-        "safety_checker": safety_checker,
-        "requires_safety_checker": requires_safety_checker,
-        "manual_seed": manual_seed,
-        "inference_steps": inference_steps,
-        "guidance_scale": guidance_scale,
-        "prompt": prompt,
-        "negative_prompt": negative_prompt,
+        "safety_checker": False,
+        "requires_safety_checker": False,
+        "manual_seed": 42,
+        "inference_steps": 10,
+        "guidance_scale": 0.5,
+        "prompt": 'A white rabbit',
+        "negative_prompt": 'lowres, cropped, worst quality, low quality, chat bubble, chat bubbles, ugly',
     }
 
+    # code output order
+    code = {}
 
-    return initial_config, devices, models, schedulers
+    code[code_pos_device] = f'device = "{device}"'
+    code[code_pos_variant] = f'variant = {initial_config['variant']}'
+    code[code_pos_tf32] = f'torch.backends.cuda.matmul.allow_tf32 = {initial_config['allow_tensorfloat32']}'
+    code[code_pos_data_type] = 'data_type = torch.bfloat16'
+    code[code_pos_init_pipeline] = 'sys.exit("No model selected!")'
+    code[code_pos_safety_checker] = 'pipeline.safety_checker = None'
+    code[code_pos_requires_safety_checker] = f'pipeline.requires_safety_checker = {initial_config['requires_safety_checker']}'
+    code[code_pos_scheduler] = 'sys.exit("No scheduler selected!")'
+    code[code_pos_generator] = f'generator = torch.Generator("{device}")'
+    code[code_pos_prompt] = f'prompt = "{initial_config['prompt']}"'
+    code[code_pos_negative_prompt] = f'negative_prompt = "{initial_config['negative_prompt']}"'
+    code[code_pos_inference_steps] = f'inference_steps = {initial_config['inference_steps']}'
+    code[code_pos_guidance_scale] = f'guidance_scale = {initial_config['guidance_scale']}'
+    code[code_pos_run_inference] = f'''image = pipeline(
+        prompt=prompt,
+        negative_prompt=negative_prompt,
+        generator=generator.manual_seed(manual_seed),
+        num_inference_steps=inference_steps, 
+        guidance_scale=guidance_scale).images[0]'''
+
+    return initial_config, devices, models, schedulers, code
 
 def init_config(request: gr.Request, inital_config):
     
@@ -110,4 +138,7 @@ def init_config(request: gr.Request, inital_config):
             return_config['manual_seed'],
             return_config['guidance_scale']
             ]
-    
+    
+def get_sorted_code():
+    
+    return '\r\n'.join(value[1] for value in sorted(code.items()))