Update app.py

app.py

@@ -7,26 +7,29 @@
 import spaces
 import os
 
-import subprocess
-subprocess.run(['sh', './conda.sh'])
+#import subprocess
+#subprocess.run(['sh', './conda.sh'])
 
-import sys
-conda_prefix = os.path.expanduser("~/miniconda3") 
-conda_bin = os.path.join(conda_prefix, "bin")
+#import sys
+#conda_prefix = os.path.expanduser("~/miniconda3") 
+#conda_bin = os.path.join(conda_prefix, "bin")
 
 # Add Conda's bin directory to your PATH
-os.environ["PATH"] = conda_bin + os.pathsep + os.environ["PATH"]
+#os.environ["PATH"] = conda_bin + os.pathsep + os.environ["PATH"]
 
 # Activate the base environment (adjust if needed)
-os.system(f'{conda_bin}/conda init  --all') 
-os.system(f'{conda_bin}/conda activate base') 
+#os.system(f'{conda_bin}/conda init  --all') 
+#os.system(f'{conda_bin}/conda activate base') 
+
+
 #os.system(f'{conda_bin}/conda install nvidia/label/cudnn-9.3.0::cudnn')
 #os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-libraries')
 #os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-libraries-dev')
 #os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-cudart')
 #os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-cudart-dev')
 #os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-nvcc')
-os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-toolkit')
+
+#os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-toolkit')
 
 #subprocess.run(['pip', 'install', 'git+https://github.com/hidet-org/hidet.git'])
 #subprocess.run(['pip', 'install', 'git+https://github.com/ford442/hidet.git@thread'])
@@ -35,7 +38,7 @@ os.system(f'{conda_bin}/conda install nvidia/label/cuda-12.4.0::cuda-toolkit')
 #os.system(f'{conda_bin}/conda install rcdr_py37::tensorrt')
 #subprocess.run(['sh', './hidet.sh'])
 #subprocess.run(['sh', './modelopt.sh'])
-import hidet
+#import hidet
 #print(dir(hidet)) 
 #import torch_tensorrt
 
@@ -54,8 +57,8 @@ import datetime
 import cyper
 from image_gen_aux import UpscaleWithModel
 import torch
-import torch._dynamo
-torch._dynamo.list_backends()
+#import torch._dynamo
+#torch._dynamo.list_backends()
 
 torch.backends.cuda.matmul.allow_tf32 = False
 torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = False
@@ -194,29 +197,29 @@ def load_and_prepare_model():
     pipe.text_encoder.eval()
     '''
 
-    pipe.unet = pipe.unet.to(memory_format=torch.contiguous_format)
+    #pipe.unet = pipe.unet.to(memory_format=torch.contiguous_format)
     
-    #pipe.unet.to(memory_format=torch.channels_last)
+    pipe.unet.to(memory_format=torch.channels_last)
     #pipe.enable_vae_tiling()
     #pipe.unet = torch.compile(pipe.unet, backend="hidet", dynamic=False, mode='max-autotune') #.to(device=device, dtype=torch.bfloat16)
     #pipe.unet = torch.compile(pipe.unet, backend="hidet", dynamic=False, mode='max-autotune-no-cudagraphs') #.to(device=device, dtype=torch.bfloat16)
     #pipe.unet = torch.compile(pipe.unet, backend="hidet", dynamic=False, options={'epilogue_fusion': True, 'shape_padding': True}) #.to(device=device, dtype=torch.bfloat16)
-    pipe.unet = torch.compile(pipe.unet, dynamic=False)
+    #pipe.unet = torch.compile(pipe.unet, dynamic=False)
     #pipe.unet = torch.compile(pipe.unet, backend="hidet", dynamic=False, options={"search_space": 0})
     #pipe.unet = torch.compile(pipe.unet, backend="torch_tensorrt", dynamic=False, options={"precision": torch.bfloat16,"optimization_level": 4,})
     pipe.to(device=device, dtype=torch.bfloat16)
 
     return pipe
 
-hidet.option.parallel_build(False)
-hidet.option.parallel_tune(2,2.0)
+#hidet.option.parallel_build(False)
+#hidet.option.parallel_tune(2,2.0)
 #torch._dynamo.config.suppress_errors = True
-torch._dynamo.disallow_in_graph(diffusers.models.attention.BasicTransformerBlock)
+#torch._dynamo.disallow_in_graph(diffusers.models.attention.BasicTransformerBlock)
 
 # more search
 #hidet.torch.dynamo_config.search_space(0)
 #hidet.torch.dynamo_config.dump_graph_ir("./local_graph")
-hidet.option.cache_dir("local_cache")
+#  hidet.option.cache_dir("local_cache")
 # automatically transform the model to use float16 data type
 #hidet.torch.dynamo_config.use_fp16(True)
 # use float16 data type as the accumulate data type in operators with reduction
@@ -229,7 +232,7 @@ hidet.option.cache_dir("local_cache")
 
 pipe = load_and_prepare_model()
 
-MAX_SEED = np.iinfo(np.int32).max
+MAX_SEED = np.iinfo(np.int64).max
 
 neg_prompt_2 = " 'non-photorealistic':1.5, 'unrealistic skin','unattractive face':1.3, 'low quality':1.1, ('dull color scheme', 'dull colors', 'digital noise':1.2),'amateurish', 'poorly drawn face':1.3, 'poorly drawn', 'distorted face', 'low resolution', 'simplistic' "