fix: rebuild for ZeroGPU torch whitelist (2.8-2.11)

ZeroGPU rejects any torch pin outside 2.8.0/2.9.1/2.10.0/2.11.0, and there
are no kaolin/pytorch3d/gsplat wheels for those versions. Instead of pinning
old torch:

- drop torch from requirements — use the ZeroGPU-provided build
- stub kaolin/pytorch3d/flash_attn (adopted from jboth/sam3d-objects-fixed);
only their pure-python surface is used since texture baking, mesh
postprocess and layout postprocess are disabled in Inference.__call__
- install spconv-cu124 (torch-independent), gsplat (PyPI JIT mode, kernels
never compiled), sam2/MoGe/utils3d at runtime against the live torch
- force sdpa attention (patch inference_pipeline.py H200 flash_attn autodetect)
- checkpoints via symlink to snapshot_download, config = pipeline.yaml
- add Diagnose tab for module/GPU health checks

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

README.md

@@ -4,8 +4,8 @@ emoji: 📦
 colorFrom: purple
 colorTo: blue
 sdk: gradio
-sdk_version: 6.0.2
-python_version: "3.10"
+sdk_version: 6.13.0
+python_version: "3.12"
 app_file: app.py
 pinned: false
 license: other

app.py

@@ -2,101 +2,209 @@
 SAM 3D Objects MCP Server
 Image → 3D Object (GLB)
 
-Automatic object detection with SAM2 + 3D reconstruction with SAM 3D Objects.
+SAM2 auto-detection + SAM 3D Objects reconstruction on ZeroGPU.
+kaolin/pytorch3d/flash_attn are stubbed (only their pure-python surface is
+used since texture baking and mesh postprocessing are disabled); spconv and
+gsplat are installed at runtime against the ZeroGPU-provided torch.
 """
 import os
 import sys
 import subprocess
 import tempfile
-import uuid
 from pathlib import Path
 
+os.environ.setdefault("CUDA_HOME", "/usr/local/cuda")
+os.environ.setdefault("CONDA_PREFIX", "/usr/local")
+os.environ["LIDRA_SKIP_INIT"] = "true"
+os.environ["ATTN_BACKEND"] = "sdpa"
+os.environ["SPARSE_ATTN_BACKEND"] = "sdpa"
+os.environ["SPARSE_BACKEND"] = "spconv"
+
+# MUST import spaces before torch
+import spaces
 import gradio as gr
 import numpy as np
-import spaces
 from huggingface_hub import snapshot_download, login
-from PIL import Image
 
-# Login with HF_TOKEN if available
 if os.environ.get("HF_TOKEN"):
-    login(token=os.environ.get("HF_TOKEN"))
-
-# Set CUDA_HOME for sam-3d-objects (expects conda but we're not using it)
-if "CUDA_HOME" not in os.environ:
-    os.environ["CUDA_HOME"] = "/usr/local/cuda"
-if "CONDA_PREFIX" not in os.environ:
-    os.environ["CONDA_PREFIX"] = "/usr/local"
-
-# Clone sam-3d-objects repo if not exists
-SAM3D_PATH = Path("/home/user/app/sam-3d-objects")
+    login(token=os.environ["HF_TOKEN"])
+
+# --- Stubs (must be on sys.path before sam3d imports) ---
+APP_ROOT = Path(__file__).parent
+for stub in ["kaolin_stub", "pytorch3d_stub", "flash_attn_stub"]:
+    stub_path = APP_ROOT / stub
+    if stub_path.exists():
+        sys.path.insert(0, str(stub_path))
+        print(f"Stub added: {stub}")
+
+
+# --- Runtime pip installs (need torch present, so not in requirements.txt) ---
+def _pip(*a):
+    r = subprocess.run(
+        [sys.executable, "-m", "pip", "install", "--no-cache-dir"] + list(a),
+        capture_output=True, text=True, timeout=1200,
+    )
+    tag = a[-1][:60]
+    print(f"  pip {'OK' if r.returncode == 0 else 'FAIL'}: {tag}")
+    if r.returncode != 0:
+        print(f"    {r.stderr[-300:]}")
+    return r.returncode == 0
+
+
+print("=== Runtime installs ===")
+_pip("open3d>=0.18.0")
+_pip("--no-deps", "git+https://github.com/EasternJournalist/utils3d.git")
+_pip("iopath")
+_pip("--no-deps", "sam2>=1.1.0")
+_pip("--no-deps", "git+https://github.com/microsoft/MoGe.git@a8c37341bc0325ca99b9d57981cc3bb2bd3e255b")
+# gsplat: no prebuilt wheels beyond pt24 — PyPI sdist installs in JIT mode;
+# kernels never compile here because rendering paths are disabled.
+_pip("--no-deps", "gsplat")
+# spconv cu124 wheel is forward-compatible with the cu128 runtime
+_pip("spconv-cu124==2.3.8")
+
+# --- Clone sam-3d-objects ---
+SAM3D_PATH = APP_ROOT / "sam-3d-objects"
 if not SAM3D_PATH.exists():
-    print("Cloning sam-3d-objects repository...")
-    subprocess.run([
-        "git", "clone",
-        "https://github.com/facebookresearch/sam-3d-objects.git",
-        str(SAM3D_PATH)
-    ], check=True)
-
-# Add both repo root and notebook folder to path
+    print("Cloning sam-3d-objects...")
+    subprocess.run(["git", "clone", "--depth", "1",
+                    "https://github.com/facebookresearch/sam-3d-objects.git",
+                    str(SAM3D_PATH)], check=True)
+
+subprocess.run([sys.executable, "-m", "pip", "install", "-e", str(SAM3D_PATH), "--no-deps"],
+               capture_output=True, text=True)
+
+hydra_patch = SAM3D_PATH / "patching" / "hydra"
+if hydra_patch.exists():
+    subprocess.run(["bash", str(hydra_patch)], capture_output=True, cwd=str(SAM3D_PATH))
+
+# Patch: inference_pipeline.py forces flash_attn on A100/H100/H200,
+# but only the stub is installed — keep sdpa.
+ip_file = SAM3D_PATH / "sam3d_objects" / "pipeline" / "inference_pipeline.py"
+if ip_file.exists():
+    src = ip_file.read_text()
+    old = ('if "A100" in gpu_name or "H100" in gpu_name or "H200" in gpu_name:\n'
+           '        # logger.info("Use flash_attn")\n'
+           '        os.environ["ATTN_BACKEND"] = "flash_attn"\n'
+           '        os.environ["SPARSE_ATTN_BACKEND"] = "flash_attn"')
+    if old in src:
+        ip_file.write_text(src.replace(
+            old,
+            '# PATCHED: flash_attn not available on ZeroGPU, keep sdpa\n'
+            '    os.environ.setdefault("ATTN_BACKEND", "sdpa")\n'
+            '    os.environ.setdefault("SPARSE_ATTN_BACKEND", "sdpa")'
+        ))
+        print("PATCHED: inference_pipeline.py — forced sdpa backend")
+    else:
+        print("INFO: inference_pipeline.py patch marker not found (upstream changed?)")
+
 sys.path.insert(0, str(SAM3D_PATH))
 sys.path.insert(0, str(SAM3D_PATH / "notebook"))
 
-# Global models
-SAM3D_MODEL = None
-SAM2_GENERATOR = None
-
-
-def load_sam2():
-    """Load SAM2 automatic mask generator"""
-    global SAM2_GENERATOR
-
-    if SAM2_GENERATOR is not None:
-        return SAM2_GENERATOR
-
-    from sam2.automatic_mask_generator import SAM2AutomaticMaskGenerator
-
-    print("Loading SAM2 model...")
-
-    SAM2_GENERATOR = SAM2AutomaticMaskGenerator.from_pretrained("facebook/sam2-hiera-large")
-
-    print("✓ SAM2 loaded")
-    return SAM2_GENERATOR
-
-
-def load_sam3d():
-    """Load SAM 3D Objects model"""
-    global SAM3D_MODEL
-
-    if SAM3D_MODEL is not None:
-        return SAM3D_MODEL
-
+# --- Patch: utils3d from git lacks depth_edge/normals_edge (needed by layout code) ---
+try:
+    import utils3d.numpy as _u3d_np
+    if not hasattr(_u3d_np, "depth_edge"):
+        def _depth_edge(depth, rtol=0.03, mask=None):
+            from scipy.ndimage import sobel
+            d = np.where(mask, depth, 0.0) if mask is not None else depth.copy()
+            grad = np.sqrt(sobel(d, axis=1) ** 2 + sobel(d, axis=0) ** 2)
+            denom = np.abs(d)
+            denom[denom < 1e-6] = 1e-6
+            edge = (grad / denom) > rtol
+            return edge & mask if mask is not None else edge
+        _u3d_np.depth_edge = _depth_edge
+
+        def _normals_edge(normals, tol=0.1, mask=None):
+            from scipy.ndimage import sobel
+            edges = np.zeros(normals.shape[:2], dtype=bool)
+            for c in range(normals.shape[-1]):
+                ch = np.where(mask, normals[..., c], 0.0) if mask is not None else normals[..., c]
+                edges |= np.sqrt(sobel(ch, axis=1) ** 2 + sobel(ch, axis=0) ** 2) > tol
+            return edges & mask if mask is not None else edges
+        _u3d_np.normals_edge = _normals_edge
+        print("Injected depth_edge + normals_edge into utils3d.numpy")
+except Exception as e:
+    print(f"utils3d patch skipped: {e}")
+
+# --- Pre-download checkpoints ---
+print("Downloading SAM3D checkpoints...")
+CKPT_DIR = snapshot_download(repo_id="facebook/sam-3d-objects",
+                             token=os.environ.get("HF_TOKEN"))
+hf_ckpt = Path(CKPT_DIR) / "checkpoints"
+local_ckpt = SAM3D_PATH / "checkpoints" / "hf"
+if hf_ckpt.exists() and not local_ckpt.exists():
+    local_ckpt.parent.mkdir(parents=True, exist_ok=True)
+    local_ckpt.symlink_to(hf_ckpt)
+CONFIG_PATH = str(local_ckpt / "pipeline.yaml")
+print(f"Config exists: {Path(CONFIG_PATH).exists()}")
+print("=== Startup complete ===")
+
+
+def _gaussians_to_glb(result, out_dir):
+    """Extract gaussians from pipeline result and Poisson-mesh them to GLB."""
     import torch
-    print("Loading SAM 3D Objects model...")
-
-    # Download checkpoints
-    checkpoint_dir = snapshot_download(
-        repo_id="facebook/sam-3d-objects",
-        token=os.environ.get("HF_TOKEN")
-    )
-
-    # Import from notebook/inference.py
-    from inference import Inference
-
-    # Config path in the repo
-    config_path = str(SAM3D_PATH / "sam3d_objects" / "configs" / "default.yaml")
-
-    SAM3D_MODEL = Inference(config_path, compile=False)
-    # Point to downloaded checkpoints
-    SAM3D_MODEL.checkpoint_dir = checkpoint_dir
-
-    print("✓ SAM 3D Objects loaded")
-    return SAM3D_MODEL
+    gs = None
+    if hasattr(result, "save_ply"):
+        gs = result
+    elif isinstance(result, dict):
+        for k in ("gs", "gaussian", "gaussians", "scene"):
+            v = result.get(k)
+            if v is not None:
+                gs = v[0] if isinstance(v, (list, tuple)) else v
+                break
+    if gs is None:
+        return None
+
+    import open3d as o3d
+    glb = f"{out_dir}/object.glb"
+    if hasattr(gs, "save_ply"):
+        ply = f"{out_dir}/temp.ply"
+        gs.save_ply(ply)
+        pcd = o3d.io.read_point_cloud(ply)
+    elif hasattr(gs, "_xyz"):
+        pcd = o3d.geometry.PointCloud()
+        pcd.points = o3d.utility.Vector3dVector(gs._xyz.detach().cpu().numpy())
+    else:
+        return None
+    pcd.estimate_normals()
+    mesh, _ = o3d.geometry.TriangleMesh.create_from_point_cloud_poisson(pcd, depth=8)
+    o3d.io.write_triangle_mesh(glb, mesh)
+    return glb
+
+
+@spaces.GPU(duration=60)
+def diagnose():
+    """Check GPU and importability of all critical modules."""
+    import torch
+    lines = [f"torch={torch.__version__}", f"cuda={torch.cuda.is_available()}"]
+    if torch.cuda.is_available():
+        lines.append(f"gpu={torch.cuda.get_device_name()}")
+    for mod in ["kaolin", "pytorch3d", "spconv", "gsplat", "utils3d", "open3d", "moge"]:
+        try:
+            m = __import__(mod)
+            lines.append(f"{mod}: OK ({getattr(m, '__version__', '-')})")
+        except Exception as e:
+            lines.append(f"{mod}: FAIL - {e}")
+    try:
+        from sam2.automatic_mask_generator import SAM2AutomaticMaskGenerator  # noqa: F401
+        lines.append("sam2: OK")
+    except Exception as e:
+        lines.append(f"sam2: FAIL - {e}")
+    try:
+        from inference import Inference  # noqa: F401
+        lines.append("SAM3D Inference: importable")
+    except Exception as e:
+        lines.append(f"SAM3D Inference: FAIL - {e}")
+    lines.append(f"config: {Path(CONFIG_PATH).exists()}")
+    return "\n".join(lines)
 
 
-@spaces.GPU(duration=120)
+@spaces.GPU(duration=300)
 def reconstruct_objects(image: np.ndarray):
     """
-    Automatically detect and reconstruct 3D objects from image.
+    Automatically detect the largest object with SAM2 and reconstruct it
+    in 3D with SAM 3D Objects.
 
     Args:
         image: Input RGB image
@@ -106,94 +214,56 @@ def reconstruct_objects(image: np.ndarray):
     """
     if image is None:
         return None, None, "❌ No image provided"
-
     try:
         import torch
-        import trimesh
-        from PIL import Image as PILImage
-
-        # Load models
-        generator = load_sam2()
-        inference = load_sam3d()
-
-        # Convert to PIL if needed
-        if isinstance(image, np.ndarray):
-            pil_image = PILImage.fromarray(image)
-        else:
-            pil_image = image
-            image = np.array(pil_image)
-
-        # Auto-detect all objects with SAM2
-        print("Detecting objects...")
-        masks = generator.generate(image)
-
-        if not masks or len(masks) == 0:
-            return None, image, "⚠️ No objects detected"
-
-        # Sort by area, take largest object
-        masks = sorted(masks, key=lambda x: x['area'], reverse=True)
-        best_mask = masks[0]['segmentation']
-
-        # Create preview with mask overlay
-        preview = image.copy()
+        import time
+        t0 = time.time()
+        print(f"GPU: {torch.cuda.get_device_name()}")
+
+        from sam2.automatic_mask_generator import SAM2AutomaticMaskGenerator
+        sam2_gen = SAM2AutomaticMaskGenerator.from_pretrained("facebook/sam2-hiera-small")
+        print(f"  SAM2 loaded ({time.time()-t0:.0f}s)")
+
+        image_np = np.array(image) if not isinstance(image, np.ndarray) else image
+        masks = sam2_gen.generate(image_np)
+        if not masks:
+            return None, image_np, "⚠️ No objects detected"
+        masks = sorted(masks, key=lambda x: x["area"], reverse=True)
+        best_mask = masks[0]["segmentation"]
+
+        preview = image_np.copy()
         preview[best_mask] = (preview[best_mask] * 0.5 + np.array([0, 255, 0]) * 0.5).astype(np.uint8)
+        print(f"  {len(masks)} masks ({time.time()-t0:.0f}s)")
 
-        # Convert mask to PIL
-        mask_pil = PILImage.fromarray((best_mask * 255).astype(np.uint8))
+        # Free SAM2 before loading SAM3D
+        del sam2_gen
+        torch.cuda.empty_cache()
 
-        # Run 3D reconstruction
-        print("Reconstructing 3D...")
-        result = inference(image=pil_image, mask=mask_pil)
+        from inference import Inference
+        sam3d = Inference(CONFIG_PATH, compile=False)
+        print(f"  SAM3D loaded ({time.time()-t0:.0f}s)")
 
+        result = sam3d(image=image_np, mask=best_mask, seed=42)
+        print(f"  Reconstructed ({time.time()-t0:.0f}s)")
         if result is None:
-            return None, preview, "⚠️ 3D reconstruction failed"
-
-        # Export as GLB
-        output_dir = tempfile.mkdtemp()
-        glb_path = f"{output_dir}/object_{uuid.uuid4().hex[:8]}.glb"
-
-        # Extract point cloud from result and convert to mesh
-        if hasattr(result, 'save_ply'):
-            # Save temp PLY then convert
-            ply_path = f"{output_dir}/temp.ply"
-            result.save_ply(ply_path)
-            # Load and convert to mesh using Open3D
-            import open3d as o3d
-            pcd = o3d.io.read_point_cloud(ply_path)
-            # Estimate normals and create mesh via Poisson reconstruction
-            pcd.estimate_normals()
-            mesh, _ = o3d.geometry.TriangleMesh.create_from_point_cloud_poisson(pcd, depth=8)
-            o3d.io.write_triangle_mesh(glb_path, mesh)
-        elif 'gaussians' in result:
-            ply_path = f"{output_dir}/temp.ply"
-            result['gaussians'].save_ply(ply_path)
-            import open3d as o3d
-            pcd = o3d.io.read_point_cloud(ply_path)
-            pcd.estimate_normals()
-            mesh, _ = o3d.geometry.TriangleMesh.create_from_point_cloud_poisson(pcd, depth=8)
-            o3d.io.write_triangle_mesh(glb_path, mesh)
-        else:
-            # Try to extract vertices
-            vertices = result.get('xyz', result.get('points', None))
-            if vertices is not None:
-                if torch.is_tensor(vertices):
-                    vertices = vertices.cpu().numpy()
-                # Create mesh from points
-                import open3d as o3d
-                pcd = o3d.geometry.PointCloud()
-                pcd.points = o3d.utility.Vector3dVector(vertices)
-                pcd.estimate_normals()
-                mesh, _ = o3d.geometry.TriangleMesh.create_from_point_cloud_poisson(pcd, depth=8)
-                o3d.io.write_triangle_mesh(glb_path, mesh)
-            else:
-                return None, preview, "⚠️ Could not extract 3D data"
-
-        return glb_path, preview, f"✓ Detected {len(masks)} objects, reconstructed largest"
+            return None, preview, "⚠️ Reconstruction returned None"
 
+        glb = _gaussians_to_glb(result, tempfile.mkdtemp())
+        if glb is None:
+            keys = list(result.keys()) if isinstance(result, dict) else dir(result)
+            return None, preview, f"⚠️ Cannot extract 3D data. Keys: {keys}"
+
+        import trimesh
+        try:
+            n_faces = len(trimesh.load(glb, force="mesh").faces)
+        except Exception:
+            n_faces = 0
+        return glb, preview, f"✓ {len(masks)} objects detected, {n_faces:,} faces ({int(time.time()-t0)}s)"
     except Exception as e:
         import traceback
-        traceback.print_exc()
-        return None, None, f"❌ Error: {e}"
+        tb = traceback.format_exc()
+        print(tb)
+        return None, None, f"❌ Error:\n{tb[-1500:]}"
 
 
 # Gradio Interface
@@ -205,27 +275,24 @@ with gr.Blocks(title="SAM 3D Objects MCP") as demo:
     Automatically detects objects and reconstructs the largest one in 3D.
     """)
 
-    with gr.Row():
-        with gr.Column():
-            input_image = gr.Image(label="Input Image", type="numpy")
-            btn = gr.Button("🚀 Detect & Reconstruct", variant="primary", size="lg")
-
-        with gr.Column():
-            preview = gr.Image(label="Detected Object", type="numpy", interactive=False)
-            status = gr.Textbox(label="Status")
-
-    with gr.Row():
-        with gr.Column():
+    with gr.Tab("Reconstruct"):
+        with gr.Row():
+            with gr.Column():
+                input_image = gr.Image(label="Input Image", type="numpy")
+                btn = gr.Button("🚀 Detect & Reconstruct", variant="primary", size="lg")
+            with gr.Column():
+                preview = gr.Image(label="Detected Object", type="numpy", interactive=False)
+                status = gr.Textbox(label="Status")
+        with gr.Row():
             output_model = gr.Model3D(label="3D Preview")
-        with gr.Column():
             output_file = gr.File(label="Download GLB")
+        btn.click(reconstruct_objects, inputs=[input_image], outputs=[output_model, preview, status])
+        output_model.change(lambda x: x, inputs=[output_model], outputs=[output_file])
 
-    btn.click(
-        reconstruct_objects,
-        inputs=[input_image],
-        outputs=[output_model, preview, status]
-    )
-    output_model.change(lambda x: x, inputs=[output_model], outputs=[output_file])
+    with gr.Tab("Diagnose"):
+        diag_btn = gr.Button("Diagnose GPU & Modules")
+        diag_out = gr.Textbox(lines=15, label="Diagnostics")
+        diag_btn.click(diagnose, outputs=[diag_out])
 
     gr.Markdown("""
     ---

flash_attn_stub/flash_attn/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/flash_attn_stub%2Fflash_attn%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fflash_attn_stub%2Fflash_attn%2F__init__.py=&etag=%22e8154a003b387b4a013f13f81e6b1d904ed91395%22

kaolin_stub/kaolin/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2F__init__.py=&etag=%22b33d127162fbf6be3b4108393add91478bb521bc%22

kaolin_stub/kaolin/ops/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2Fops%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2Fops%2F__init__.py=&etag=%227f1d5481770b3ffa39effdbf052adca321305592%22

kaolin_stub/kaolin/ops/mesh.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2Fops%2Fmesh.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2Fops%2Fmesh.py=&etag=%22d225c551eeaf6ff221c55bdc3521cb0a49ba8db8%22

kaolin_stub/kaolin/render/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2Frender%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2Frender%2F__init__.py=&etag=%228b7c03b43ec0b7f5d8fa5812a970190761e6ec80%22

kaolin_stub/kaolin/render/camera.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2Frender%2Fcamera.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2Frender%2Fcamera.py=&etag=%2297e63314864f4ef5e2d5d3a21971d3a9297ebee3%22

kaolin_stub/kaolin/utils/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2Futils%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2Futils%2F__init__.py=&etag=%2297df528c28d988591233bdc1ef2ba44e78cddf94%22

kaolin_stub/kaolin/utils/testing.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2Futils%2Ftesting.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2Futils%2Ftesting.py=&etag=%2270331aa66fe7f80563073a85d5102d6485e7213f%22

kaolin_stub/kaolin/visualize/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/kaolin_stub%2Fkaolin%2Fvisualize%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fkaolin_stub%2Fkaolin%2Fvisualize%2F__init__.py=&etag=%228492fef06d88908029b877d623ab3f9ec746a767%22

pytorch3d_stub/pytorch3d/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2F__init__.py=&etag=%220be164bba8a444d77ed028c703c53164efa59be1%22

pytorch3d_stub/pytorch3d/renderer/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Frenderer%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Frenderer%2F__init__.py=&etag=%224ad1da5bd5d463828d7c2da904e7e130b1644f44%22

pytorch3d_stub/pytorch3d/renderer/camera_utils.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Frenderer%2Fcamera_utils.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Frenderer%2Fcamera_utils.py=&etag=%22819d1ca667d03548de636a7079ab25e239735d6c%22

pytorch3d_stub/pytorch3d/renderer/cameras.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Frenderer%2Fcameras.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Frenderer%2Fcameras.py=&etag=%2222da58b8bd463e8347effc1530fac0e77313a845%22

pytorch3d_stub/pytorch3d/renderer/mesh/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Frenderer%2Fmesh%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Frenderer%2Fmesh%2F__init__.py=&etag=%2254776a5dc4978bdd6adc39d981fe242aaa0327f7%22

pytorch3d_stub/pytorch3d/renderer/mesh/textures.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Frenderer%2Fmesh%2Ftextures.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Frenderer%2Fmesh%2Ftextures.py=&etag=%2254776a5dc4978bdd6adc39d981fe242aaa0327f7%22

pytorch3d_stub/pytorch3d/structures/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Fstructures%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Fstructures%2F__init__.py=&etag=%221d9561d58a998fb884e75446b8a8417d346bfa9a%22

pytorch3d_stub/pytorch3d/transforms/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Ftransforms%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Ftransforms%2F__init__.py=&etag=%22561aad8e8e0e3e6182e3f62283f3cbdd710e4bf8%22

pytorch3d_stub/pytorch3d/vis/__init__.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Fvis%2F__init__.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Fvis%2F__init__.py=&etag=%22c887a2b86e91aa7227f07cf6d66d964ee465d552%22

pytorch3d_stub/pytorch3d/vis/plotly_vis.py

@@ -0,0 +1 @@
+Temporary Redirect. Redirecting to /api/resolve-cache/spaces/jboth/sam3d-objects-fixed/7b8ab13b284b67e82dd8c321d957c0080dd66a51/pytorch3d_stub%2Fpytorch3d%2Fvis%2Fplotly_vis.py?%2Fspaces%2Fjboth%2Fsam3d-objects-fixed%2Fresolve%2Fmain%2Fpytorch3d_stub%2Fpytorch3d%2Fvis%2Fplotly_vis.py=&etag=%22623e37cd1073af06175464439f26d85ed2b79241%22

requirements.txt

@@ -1,25 +1,50 @@
---extra-index-url https://download.pytorch.org/whl/cu121
---extra-index-url https://docs.gsplat.studio/whl/pt25cu121
---find-links https://nvidia-kaolin.s3.us-east-2.amazonaws.com/torch-2.5.1_cu121.html
-torch==2.5.1+cu121
-torchvision==0.20.1+cu121
-gradio>=6.0.2
-huggingface_hub>=0.26.0
-spaces>=0.30.0
-numpy>=1.26.0
-opencv-python>=4.8.0
-Pillow>=10.0.0
-trimesh>=4.0.0
-open3d>=0.18.0
-hydra-core>=1.3.0
-omegaconf>=2.3.0
-einops>=0.7.0
-loguru>=0.7.0
-seaborn==0.13.2
-ninja
+numpy
+Pillow
+trimesh
+hydra-core
+hydra-submitit-launcher
+omegaconf
+einops
+einops-exts
+loguru
+seaborn
 jaxtyping
 rich
-kaolin==0.17.0
-gsplat
-sam2>=1.1.0
-utils3d
+torchvision
+opencv-python-headless
+ninja
+timm
+optree
+roma
+fvcore
+rootutils
+easydict
+scikit-image
+colorama
+ftfy
+astor
+av
+gdown
+deprecation
+orjson
+randomname
+simplejson
+pydot
+igraph
+plyfile
+point-cloud-utils
+pymeshfix
+xatlas
+pycocotools
+lightning
+pytorch-lightning
+scipy
+pyvista
+plotly
+kaleido
+webdataset
+wandb
+tensorboard
+sentence-transformers
+accelerate
+safetensors