Convert to cog.

.gitignore

@@ -0,0 +1,2 @@
+
+render.obj

README.md

@@ -10,4 +10,16 @@ pinned: false
 license: mit
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+```bash
+cog predict -i text="cat"
+```
+
+Currently, this project is in the **Sandbox** phase, which is the initial stage of a project according to the V-Sekai maturity model.
+
+The V-Sekai maturity model consists of three stages:
+
+1. **Sandbox**: This stage serves as the starting point for projects, aiming to nurture innovation and project development.
+2. **Incubating**: At this stage, projects are on their journey towards graduation, with an expectation of having a stable codebase and well-structured formation.
+3. **Graduated**: These projects have reached maturity, adopting best practices and catering to a broad user base.
+
+As the project fulfills the criteria laid out by V-Sekai, it will progress through these stages.

app.py

@@ -1,49 +0,0 @@
-import torch  
-import time
-from meshgpt_pytorch import ( 
-    MeshTransformer,
-    mesh_render
-)
-import igl
-
-import gradio as gr
-import numpy as np 
-
-transformer = MeshTransformer.from_pretrained("MarcusLoren/MeshGPT-preview")
-def save_as_obj(file_path):
-    v, f = igl.read_triangle_mesh(file_path)
-    v, f, _, _ = igl.remove_unreferenced(v, f)
-    c, _ = igl.orientable_patches(f)
-    f, _ = igl.orient_outward(v, f, c)
-    igl.write_triangle_mesh(file_path, v, f)
-    return file_path
-
-def predict(text, num_input, num_temp):
-    transformer.eval()
-    labels = [label.strip() for label in text.split(',')] 
-    output = []
-    current_time = time.time() 
-    formatted_time = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(current_time))
-    print(formatted_time, " Input:", text, "num_input", num_input, "num_temp",num_temp)
-    if num_input > 1:
-        for label in labels:
-            output.append((transformer.generate(texts = [label ] * num_input, temperature = num_temp)))   
-    else:
-        output.append((transformer.generate(texts = labels  , temperature = num_temp)))   
-            
-    mesh_render.save_rendering('./render.obj', output) 
-    return save_as_obj('./render.obj')
-
-gradio_app = gr.Interface(
-    predict, 
-    inputs=[
-        gr.Textbox(label="Enter labels, separated by commas"),
-        gr.Number(value=1, label="Number of examples per input"),
-        gr.Slider(minimum=0, maximum=1, value=0, label="Temperature (0 to 1)")
-    ],
-    outputs=gr.Model3D(clear_color=[0.0, 0.0, 0.0, 0.0], label="3D Model"),
-    title="MeshGPT Inference - (Rendering doesn't work, please download for best result)",
-)
-
-if __name__ == "__main__":
-    gradio_app.launch(share=False)

cog.yaml

@@ -0,0 +1,12 @@
+build:
+  gpu: true
+  system_packages:
+    - "libgl1-mesa-glx"
+    - "libglib2.0-0"
+  python_version: "3.11"
+  cuda: "11.5" # Query nvcc --version
+  python_packages:
+    - "torch==2.3.1"
+    - "libigl==2.5.1"
+    - "git+https://github.com/MarcusLoppe/meshgpt-pytorch.git@14126a20626356e9bd2e12dcf1a06872b9fe968d"
+predict: "predict.py:Predictor"

predict.py

@@ -0,0 +1,72 @@
+from cog import BasePredictor, Input, File
+import torch
+import time
+from meshgpt_pytorch import MeshTransformer, mesh_render
+import igl
+import numpy as np
+
+class Predictor(BasePredictor):
+    def setup(self):
+        """Load the model into memory to make running multiple predictions efficient"""
+        self.transformer = MeshTransformer.from_pretrained(
+            "MarcusLoren/MeshGPT-preview"
+        )
+
+    def save_as_obj(self, file_path):
+        v, f = igl.read_triangle_mesh(file_path)
+        v, f, _, _ = igl.remove_unreferenced(v, f)
+        c, _ = igl.orientable_patches(f)
+        f, _ = igl.orient_outward(v, f, c)
+        igl.write_triangle_mesh(file_path, v, f)
+        return file_path
+
+    def predict(
+        self,
+        text: str = Input(description="Enter labels, separated by commas"),
+        num_input: int = Input(description="Number of examples per input", default=1),
+        num_temp: float = Input(description="Temperature (0 to 1)", default=0),
+    ) -> File:
+        """Run a single prediction on the model"""
+        self.transformer.eval()
+        labels = [label.strip() for label in text.split(",")]
+
+        output = []
+        current_time = time.time()
+
+        formatted_time = time.strftime(
+            "%Y-%m-%d %H:%M:%S", time.localtime(current_time)
+        )
+        print(
+            formatted_time,
+            " Input:",
+            text,
+            "num_input",
+            num_input,
+            "num_temp",
+            num_temp,
+        )
+
+        if num_input > 1:
+            for label in labels:
+                output.append(
+                    (
+                        self.transformer.generate(
+                            texts=[label] * num_input, temperature=num_temp
+                        )
+                    )
+                )
+        else:
+            output.append(
+                (self.transformer.generate(texts=labels, temperature=num_temp))
+            )
+
+        mesh_render.save_rendering("./render.obj", output)
+
+        return self.save_as_obj("./render.obj")
+
+
+if __name__ == "__main__":
+    predictor = Predictor()
+    predictor.setup()
+
+    print(predictor.predict("cat", 1, 0.0))

requirements.txt

@@ -1,2 +0,0 @@
- git+https://github.com/MarcusLoppe/meshgpt-pytorch.git
- libigl