Implement temporary directory cleanup every 10 minutes and remove unused kaolin imports

app.py

@@ -309,6 +309,22 @@ with gr.Blocks(title="Game Items Generator") as demo:
     )
     
 
+# Cleans up the temporary directory every 10 minutes
+import threading
+import time
+
+def cleanup_tmp_dir():
+    while True:
+        if os.path.exists(TMP_DIR):
+            for file in os.listdir(TMP_DIR):
+                # remove files older than 10 minutes
+                if time.time() - os.path.getmtime(os.path.join(TMP_DIR, file)) > 600:
+                    os.remove(os.path.join(TMP_DIR, file))
+        time.sleep(600)
+                
+cleanup_thread = threading.Thread(target=cleanup_tmp_dir)
+cleanup_thread.start()
+    
 # Launch the Gradio app
 if __name__ == "__main__":
     pipeline = TrellisImageTo3DPipeline.from_pretrained("JeffreyXiang/TRELLIS-image-large")

requirements.txt

@@ -1,7 +1,5 @@
 --extra-index-url https://download.pytorch.org/whl/cu121
---find-links https://nvidia-kaolin.s3.us-east-2.amazonaws.com/torch-2.4.0_cu121.html
 
-spaces
 torch==2.4.0
 torchvision==0.19.0
 pillow==10.4.0
@@ -20,10 +18,8 @@ pymeshfix==0.17.0
 igraph==0.11.8
 git+https://github.com/EasternJournalist/utils3d.git@9a4eb15e4021b67b12c460c7057d642626897ec8
 xformers==0.0.27.post2
-kaolin==0.17.0
 spconv-cu120==2.3.6
 transformers==4.46.3
-gradio_litmodel3d==0.0.1
 https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.0.post2/flash_attn-2.7.0.post2+cu12torch2.4cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
 https://huggingface.co/spaces/JeffreyXiang/TRELLIS/resolve/main/wheels/diff_gaussian_rasterization-0.0.0-cp310-cp310-linux_x86_64.whl?download=true
 https://huggingface.co/spaces/JeffreyXiang/TRELLIS/resolve/main/wheels/nvdiffrast-0.3.3-cp310-cp310-linux_x86_64.whl?download=true

trellis/renderers/mesh_renderer.py

@@ -6,11 +6,6 @@
 # distribution of this software and related documentation without an express
 # license agreement from NVIDIA CORPORATION & AFFILIATES is strictly prohibited.
 import torch
-try:
-    import kaolin as kal
-    import nvdiffrast.torch as dr
-except :
-    print("Kaolin and nvdiffrast are not installed. Please install them to use the mesh renderer.")
 from easydict import EasyDict as edict
 from ..representations.mesh import MeshExtractResult
 import torch.nn.functional as F

trellis/representations/mesh/cube2mesh.py

@@ -9,10 +9,6 @@ import torch
 from ...modules.sparse import SparseTensor
 from easydict import EasyDict as edict
 from .utils_cube import *
-try:
-    from .flexicube import FlexiCubes
-except:
-    print("Please install kaolin and diso to use the mesh extractor.")
 
 
 class MeshExtractResult:

trellis/representations/mesh/flexicube.py

@@ -8,7 +8,6 @@
 
 import torch
 from .tables import *
-from kaolin.utils.testing import check_tensor
 
 __all__ = [
     'FlexiCubes'
@@ -49,29 +48,6 @@ class FlexiCubes:
 
     def __call__(self, voxelgrid_vertices, scalar_field, cube_idx, resolution, qef_reg_scale=1e-3,
                  weight_scale=0.99, beta=None, alpha=None, gamma_f=None, voxelgrid_colors=None, training=False):
-        assert torch.is_tensor(voxelgrid_vertices) and \
-            check_tensor(voxelgrid_vertices, (None, 3), throw=False), \
-            "'voxelgrid_vertices' should be a tensor of shape (num_vertices, 3)"
-        num_vertices = voxelgrid_vertices.shape[0]
-        assert torch.is_tensor(scalar_field) and \
-            check_tensor(scalar_field, (num_vertices,), throw=False), \
-            "'scalar_field' should be a tensor of shape (num_vertices,)"
-        assert torch.is_tensor(cube_idx) and \
-            check_tensor(cube_idx, (None, 8), throw=False), \
-            "'cube_idx' should be a tensor of shape (num_cubes, 8)"
-        num_cubes = cube_idx.shape[0]
-        assert beta is None or (
-            torch.is_tensor(beta) and
-            check_tensor(beta, (num_cubes, 12), throw=False)
-        ), "'beta' should be a tensor of shape (num_cubes, 12)"
-        assert alpha is None or (
-            torch.is_tensor(alpha) and
-            check_tensor(alpha, (num_cubes, 8), throw=False)
-        ), "'alpha' should be a tensor of shape (num_cubes, 8)"
-        assert gamma_f is None or (
-            torch.is_tensor(gamma_f) and
-            check_tensor(gamma_f, (num_cubes,), throw=False)
-        ), "'gamma_f' should be a tensor of shape (num_cubes,)"
 
         surf_cubes, occ_fx8 = self._identify_surf_cubes(scalar_field, cube_idx)
         if surf_cubes.sum() == 0: