Upload folder using huggingface_hub

app.py

@@ -14,7 +14,9 @@ MAX_SEED = np.iinfo(np.int32).max
 TMP_DIR = os.path.join(current_path, 'out')
 os.makedirs(TMP_DIR, exist_ok=True)
 TAG = "gsdiff_gobj83k_sd15__render"
-# subprocess.run(shlex.split("pip install extensions/RaDe-GS/submodules/diff-gaussian-rasterization/"))
+os.system("cd /home/paulpanwang/workspace/DiffSplat/extensions/RaDe-GS/submodules && pip3 install diff-gaussian-rasterization")
+os.system("cd /home/paulpanwang/workspace/DiffSplat/extensions/RaDe-GS/submodules/diff-gaussian-rasterization && python3 setup.py bdist_wheel ")
+
 
 # download checkpoints
 # subprocess.run(shlex.split("python3 download_ckpt.py --model_type pas")) # for txt condition
@@ -31,7 +33,7 @@ TAG = "gsdiff_gobj83k_sd15__render"
 # SD1.5
 subprocess.run(shlex.split("python3 download_ckpt.py --model_type sd15")) # for txt condition
 # subprocess.run(shlex.split("python3 download_ckpt.py --model_type sd15 --image_cond")) # for img condition
-img_commands = "PYTHONPATH=./ bash scripts/infer.sh src/infer_gsdiff_sd.py configs/gsdiff_sd15.yaml  \
+img_commands = "PYTHONPATH=./ bash scripts/infer.sh src/infer_gsdiff_sd.py configs/gsdiff_sd15.yaml {} \
 --rembg_and_center --triangle_cfg_scaling --save_ply --output_video_type mp4 --guidance_scale {} \
 --image_path {} --elevation {} --prompt {} --seed {}"
 
@@ -43,11 +45,11 @@ txt_commands = "PYTHONPATH=./ bash scripts/infer.sh src/infer_gsdiff_sd.py  conf
 
 # process function
 @spaces.GPU
-def process(input_image, prompt='a_high_quality_3D_asset', prompt_neg='ugly, blurry, pixelated obscure, unnatural colors, poor lighting, dull, unclear, cropped, lowres, low quality, artifacts, duplicate', input_elevation=20, guidance_scale=2., input_seed=0):
+def process(input_image, prompt='', prompt_neg='ugly, blurry, pixelated obscure, unnatural colors, poor lighting, dull, unclear, cropped, lowres, low quality, artifacts, duplicate', input_elevation=20, guidance_scale=2., input_seed=0):
 
     if input_image is not None:
         image_path = os.path.join(TMP_DIR, "input_image.png")
-        image_name = image_path.split('/')[-1].split('.')[0]
+        image_name = image_path.split('/')[-1].split('.')[0] + "_rgba"
         input_image.save(image_path)
         full_command = img_commands.format(TAG, guidance_scale, image_path, input_elevation, prompt, input_seed)
     else:
@@ -106,7 +108,7 @@ with block:
             input_text = gr.Textbox(label="prompt",value="a_high_quality_3D_asset")
             
             # negative prompt
-            input_neg_text = gr.Textbox(label="negative prompt", value="")
+            input_neg_text = gr.Textbox(label="negative prompt", value="ugly, blurry, pixelated obscure, unnatural colors, poor lighting, dull, unclear, cropped, lowres, low quality, artifacts, duplicate")
 
             # guidance_scale
             guidance_scale = gr.Slider(label="guidance scale", minimum=1., maximum=7.5, step=0.5, value=2.0)

requirements.txt

@@ -29,5 +29,4 @@ spaces
 triton
 wandb
 opencv-python==4.11.0.86
-https://huggingface.co/spaces/JeffreyXiang/TRELLIS/resolve/main/wheels/diff_gaussian_rasterization-0.0.0-cp310-cp310-linux_x86_64.whl?download=true
 

src/infer_gsdiff_sd.py

@@ -347,13 +347,13 @@ def main():
         level=logging.INFO
     )
     logger = logging.getLogger(__name__)
-    file_handler = logging.FileHandler(os.path.join(args.output_dir, args.tag, "log_infer.txt"))  # output to file
-    file_handler.setFormatter(logging.Formatter(
-        fmt="%(asctime)s - %(message)s",
-        datefmt="%Y/%m/%d %H:%M:%S"
-    ))
-    logger.addHandler(file_handler)
-    logger.propagate = True  # propagate to the root logger (console)
+    # file_handler = logging.FileHandler(os.path.join(args.output_dir, args.tag, "log_infer.txt"))  # output to file
+    # file_handler.setFormatter(logging.Formatter(
+    #     fmt="%(asctime)s - %(message)s",
+    #     datefmt="%Y/%m/%d %H:%M:%S"
+    # ))
+    # logger.addHandler(file_handler)
+    # logger.propagate = True  # propagate to the root logger (console)
 
     # Set the random seed
     if args.seed >= 0:
@@ -368,7 +368,7 @@ def main():
     if (args.image_path is not None or args.image_dir is not None) and args.load_pretrained_controlnet is None:
         opt.prediction_type = "v_prediction"
         opt.view_concat_condition = True
-        opt.input_concat_binary_mask = True
+        opt.input_concat_binary_mask = False
         if args.guidance_scale > 3.:
             logger.info(
                 f"WARNING: guidance scale ({args.guidance_scale}) is too large for image-conditioned models. " +