delete most cpu speed up methods, most make no difference

README.md

@@ -8,10 +8,10 @@ sdk_version: 4.24.0
 app_file: app.py
 pinned: false
 license: cc-by-nc-4.0
-# preload_from_hub:
-#     - SG161222/Realistic_Vision_V4.0_noVAE
-#     - stabilityai/sd-vae-ft-mse
-#     - runwayml/stable-diffusion-v1-5 scheduler/scheduler_config.json
+preload_from_hub:
+    - SG161222/Realistic_Vision_V4.0_noVAE Realistic_Vision_V4.0.safetensors
+    - stabilityai/sd-vae-ft-mse
+    - runwayml/stable-diffusion-v1-5 scheduler/scheduler_config.json
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -2,19 +2,10 @@
 import os
 from os.path import join as opj
 
-LD_PRELOAD = os.getenv("LD_PRELOAD")
-os.environ["LD_PRELOAD"] = f"{LD_PRELOAD}:/usr/lib/x86_64-linux-gnu/libjemalloc.so"
-os.environ["MALLOC_CONF"] = "oversize_threshold:1,background_thread:true,metadata_thp:auto,dirty_decay_ms: 60000,muzzy_decay_ms:60000"
-
-LD_PRELOAD= os.getenv("LD_PRELOAD")
-os.environ["LD_PRELOAD"] = f"{LD_PRELOAD}:/usr/lib/x86_64-linux-gnu/libjemalloc.so:/usr/lib/x86_64-linux-gnu/libiomp5.so"
-os.environ["OMP_NUM_THREADS"] ==32
-
 token = os.getenv("ACCESS_TOKEN")
 os.system(f"python -m pip install git+https://{token}@github.com/logn-2024/StableGarment.git")
 
 import torch
-import spaces
 import gradio as gr
 from PIL import Image
 import numpy as np
@@ -44,11 +35,11 @@ pretrained_garment_encoder_path = "loooooong/StableGarment_text2img"
 garment_encoder = GarmentEncoderModel.from_pretrained(pretrained_garment_encoder_path,torch_dtype=torch_dtype,subfolder="garment_encoder")
 garment_encoder = garment_encoder.to(device=device,dtype=torch_dtype)
 
-pipeline_t2i = StableGarmentPipeline.from_pretrained(base_model_path, vae=vae, torch_dtype=torch_dtype,).to(device=device) #  variant="fp16"
+pipeline_t2i = StableGarmentPipeline.from_pretrained(base_model_path, vae=vae, torch_dtype=torch_dtype, use_safetensors=True,).to(device=device) #  variant="fp16"
 # pipeline = StableDiffusionPipeline.from_pretrained("SG161222/Realistic_Vision_V4.0_noVAE", vae=vae, torch_dtype=torch_dtype).to(device=device)
 pipeline_t2i.scheduler = scheduler
 
-if False: #device=="cpu":
+if device=="cpu":
     # speed up for cpu
     # to channels last
     pipeline_t2i.unet = pipeline_t2i.unet.to(memory_format=torch.channels_last)
@@ -100,7 +91,6 @@ def prepare_controlnet_inputs(agn_mask_list,densepose_list):
     controlnet_inputs = torch.stack(controlnet_inputs)
     return controlnet_inputs
 
-@spaces.GPU(enable_queue=True)
 def tryon(prompt,init_image,garment_top,garment_down,):
     basename = os.path.splitext(os.path.basename(init_image))[0]
     image_agn = Image.open(opj(parse_dir,basename+"_agn.jpg")).resize((width,height))
@@ -120,7 +110,6 @@ def tryon(prompt,init_image,garment_top,garment_down,):
                   generator=None,).images
     return images[0]
 
-@spaces.GPU(enable_queue=True)
 def text2image(prompt,init_image,garment_top,garment_down,style_fidelity=1.):
 
     garment_top = Image.open(garment_top).resize((width,height))

packages.txt

@@ -1,2 +0,0 @@
-libjemalloc-dev
-intel-mkl