before_code stringlengths 30 149k | reviewer_comment stringlengths 16 20.6k | after_code stringlengths 30 163k | diff_context stringlengths 0 22.4k | file_path stringlengths 6 155 | comment_line int64 0 26 | language stringclasses 19
values | quality_score float64 0.07 1 | comment_type stringclasses 9
values | comment_length int64 16 20.6k | before_lines int64 2 4.61k | after_lines int64 2 4.4k | is_negative bool 2
classes | pr_title stringlengths 7 140 | pr_number int64 18 61.2k | repo_name stringclasses 29
values | repo_stars int64 321 56k | repo_language stringclasses 14
values | reviewer_username stringclasses 378
values | author_username stringclasses 739
values | user stringlengths 361 200k | assistant stringlengths 16 20.6k | system stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
"""
scene, liquid = sph_scene
sph_solver = scene.sph_solver
n_particles = sph_solver.n_particles
# Get initial positions for comparison
initial_pos = liquid.get_particles_pos().clone()
# Run simulation for a few steps
num_steps = 5
max_velocities = []
for step in range(num_ste... | Remove these comments, they do not bring any value. It is already clear from the assert itself. |
A stable simulation should:
1. Not produce NaN or Inf values in particle positions/velocities
2. Have bounded maximum velocities (not exploding)
3. Particles should remain within the simulation domain
This test runs a few simulation steps and verifies stability.
"""
scene, liquid = sph_sce... | @@ -0,0 +1,302 @@
+# Tests for SPH simulation - initial density and pressure validation
+#
+# Background:
+# SPH is sensitive to the initial particle distribution, as it directly determines the initial density and pressure fields.
+# To ensure numerical stability, particles must be initialized using a regular sampler t... | tests/test_sph.py | 26 | Python | 0.286 | suggestion | 95 | 51 | 51 | false | [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues. | 2,280 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | erizmr | Review this pull request change.
Language: Python
File: tests/test_sph.py
PR Title: [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues.
BEFORE CODE: """
scene, liquid = sph_scene
sph_solver = scene.sph_solver
n_particles = sph_solver.n_particles
# Get initial position... | Remove these comments, they do not bring any value. It is already clear from the assert itself. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
vel = liquid.get_particles_vel()
# Check for NaN/Inf in positions
assert not pos.isnan().any(), f"NaN values detected in positions at step {step}"
assert not pos.isinf().any(), f"Inf values detected in positions at step {step}"
# Check for NaN/Inf in velocities
assert n... | Remove `.item()` it is useless here. | max_velocities = []
for step in range(num_steps):
scene.step()
pos = liquid.get_particles_pos()
vel = liquid.get_particles_vel()
assert not pos.isnan().any(), f"NaN values detected in positions at step {step}"
assert not pos.isinf().any(), f"Inf values detected in positio... | @@ -0,0 +1,302 @@
+# Tests for SPH simulation - initial density and pressure validation
+#
+# Background:
+# SPH is sensitive to the initial particle distribution, as it directly determines the initial density and pressure fields.
+# To ensure numerical stability, particles must be initialized using a regular sampler t... | tests/test_sph.py | 26 | Python | 0.286 | suggestion | 36 | 51 | 51 | false | [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues. | 2,280 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | erizmr | Review this pull request change.
Language: Python
File: tests/test_sph.py
PR Title: [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues.
BEFORE CODE: vel = liquid.get_particles_vel()
# Check for NaN/Inf in positions
assert not pos.isnan().any(), f"NaN values detect... | Remove `.item()` it is useless here. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
n_particles = sph_solver.n_particles
# Get initial positions for comparison
initial_pos = liquid.get_particles_pos().clone()
# Run simulation for a few steps
num_steps = 5
max_velocities = []
for step in range(num_steps):
scene.step()
# Get current state
pos = liq... | You can remove `.item()` here I think. | 2. Have bounded maximum velocities (not exploding)
3. Particles should remain within the simulation domain
This test runs a few simulation steps and verifies stability.
"""
scene, liquid = sph_scene
sph_solver = scene.sph_solver
n_particles = sph_solver.n_particles
# Get initial positi... | @@ -0,0 +1,302 @@
+# Tests for SPH simulation - initial density and pressure validation
+#
+# Background:
+# SPH is sensitive to the initial particle distribution, as it directly determines the initial density and pressure fields.
+# To ensure numerical stability, particles must be initialized using a regular sampler t... | tests/test_sph.py | 26 | Python | 0.286 | suggestion | 38 | 51 | 51 | false | [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues. | 2,280 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | erizmr | Review this pull request change.
Language: Python
File: tests/test_sph.py
PR Title: [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues.
BEFORE CODE: n_particles = sph_solver.n_particles
# Get initial positions for comparison
initial_pos = liquid.get_particles_pos().clone()
... | You can remove `.item()` here I think. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
regardless of particle size, as the SPH kernel is scaled accordingly.
"""
scene = gs.Scene(
sim_options=gs.options.SimOptions(
dt=4e-3,
substeps=10,
),
sph_options=gs.options.SPHOptions(
lower_bound=(-0.5, -0.5, 0.0),
upper_bound=(0.5, ... | This could be unified with your `sph_scene` fixture? Maybe add some function `create_sph_scene` and you call with some default arguments in fixture, or you can call manually without non-default argument in needed? | regardless of particle size, as the SPH kernel is scaled accordingly.
"""
scene, _ = create_sph_scene(show_viewer, particle_size=particle_size)
sph_solver = scene.sph_solver
n_particles = sph_solver.n_particles
rho0 = sph_solver.particles_info[0].rho
# Get density values
densities, pre... | @@ -0,0 +1,302 @@
+# Tests for SPH simulation - initial density and pressure validation
+#
+# Background:
+# SPH is sensitive to the initial particle distribution, as it directly determines the initial density and pressure fields.
+# To ensure numerical stability, particles must be initialized using a regular sampler t... | tests/test_sph.py | 26 | Python | 0.643 | question | 213 | 51 | 31 | false | [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues. | 2,280 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | erizmr | Review this pull request change.
Language: Python
File: tests/test_sph.py
PR Title: [BUG FIX] Fix wrong sampler for SPH solver causing numerical stability issues.
BEFORE CODE: regardless of particle size, as the SPH kernel is scaled accordingly.
"""
scene = gs.Scene(
sim_options=gs.options.SimOp... | This could be unified with your `sph_scene` fixture? Maybe add some function `create_sph_scene` and you call with some default arguments in fixture, or you can call manually without non-default argument in needed? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._shared_metadata.renderer.build()
self._shared_metadata.sensors.append(self)
# Add lights from options to the shared metadata
for light_config in self._options.lights:
if self._shared_metadata.lights is not None:
# Convert light config to rasterizer... | why is it TODO? Isn't it the line above already removing camera? | options: RasterizerCameraOptions,
idx: int,
data_cls: Type[CameraData],
manager: "gs.SensorManager",
):
super().__init__(options, idx, data_cls, manager)
self._options: RasterizerCameraOptions
self._camera_node = None
self._camera_target = None
... | @@ -427,6 +427,23 @@ def build(self):
w, h = self._options.res
self._shared_metadata.image_cache[self._idx] = torch.zeros((_B, h, w, 3), dtype=torch.uint8, device=gs.device)
+ def destroy(self):
+ if not self._shared_metadata.sensors:
+ return
+ self._shared_metadata.sens... | genesis/engine/sensors/camera.py | 26 | Python | 0.214 | question | 64 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | nimrod-gileadi | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: self._shared_metadata.renderer.build()
self._shared_metadata.sensors.append(self)
# Add lights from options to ... | why is it TODO? Isn't it the line above already removing camera? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
fov=self._options.fov,
aperture=self._options.aperture,
focus_dist=self._options.focus_dist,
GUI=False,
spp=self._options.spp,
denoise=self._options.denoise,
near=0.05,
far=100.0,
env_idx=None if n_envs == 0 else... | what could be the other cleanups? Any known issue? | elif self._link is not None:
# Link exists but not built yet - use configured pose as-is (treat as world coordinates for now)
# This will be corrected when move_to_attach is called
pass
self._camera_obj = visualizer.add_camera(
res=self._options.res,
... | @@ -631,6 +648,12 @@ def build(self):
w, h = self._options.res
self._shared_metadata.image_cache[self._idx] = torch.zeros((_B, h, w, 3), dtype=torch.uint8, device=gs.device)
+ def destroy(self):
+ if not self._shared_metadata.sensors:
+ return
+ self._shared_metadata.sens... | genesis/engine/sensors/camera.py | 26 | Python | 0.286 | question | 50 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | nimrod-gileadi | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: fov=self._options.fov,
aperture=self._options.aperture,
focus_dist=self._options.focus_dist,
... | what could be the other cleanups? Any known issue? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
for sensor in sensors:
update_ground_truth_only &= sensor._options.update_ground_truth_only
sensor._cache_idx = cache_size_per_dtype[dtype]
cache_size_per_dtype[dtype] += sensor._cache_size
max_buffer_len = max(max_buffer_len, sensor._delay_ts ... | Should the destroy() method be added to BaseSensor as an default no-op method to avoid the hasattr check? | for sensor in sensors:
update_ground_truth_only &= sensor._options.update_ground_truth_only
sensor._cache_idx = cache_size_per_dtype[dtype]
cache_size_per_dtype[dtype] += sensor._cache_size
max_buffer_len = max(max_buffer_len, sensor._delay_ts ... | @@ -74,6 +74,13 @@ def build(self):
sensor.build()
sensor._is_built = True
+ def destroy(self):
+ for _sensor_cls, sensors in self._sensors_by_type.items():
+ for sensor in sensors:
+ if hasattr(sensor, "destroy"): | genesis/engine/sensors/sensor_manager.py | 26 | Python | 0.429 | question | 105 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | nimrod-gileadi | Review this pull request change.
Language: Python
File: genesis/engine/sensors/sensor_manager.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: for sensor in sensors:
update_ground_truth_only &= sensor._options.update_ground_truth_only
... | Should the destroy() method be added to BaseSensor as an default no-op method to avoid the hasattr check? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
entity_idx=sphere.idx,
link_idx_local=0,
offset_T=trans_to_T(np.array([0.0, 0.0, 3.0])),
)
)
scene.build(n_envs=n_envs)
scene.step()
for camera in (camera_1, camera_2):
data = camera.read()
if n_envs > 0:
for i in range(n_envs):
... | I think it should "crash" actually. In a sense that only a scene that was build should be destroyed. Here it should raise an exception. Of course, no rendering backend should be started before build. | **camera_common_options,
entity_idx=sphere.idx,
link_idx_local=0,
offset_T=trans_to_T(np.array([0.0, 0.0, 3.0])),
)
)
scene.build(n_envs=n_envs)
scene.step()
for camera in (camera_1, camera_2):
data = camera.read()
if n_envs > 0:
... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.571 | bug | 199 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: entity_idx=sphere.idx,
link_idx_local=0,
offset_T=trans_to_T(np.array([0.0, 0.0, 3.0])),
)
)
... | I think it should "crash" actually. In a sense that only a scene that was build should be destroyed. Here it should raise an exception. Of course, no rendering backend should be started before build. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
@pytest.mark.required
def test_destroy_unbuilt_scene_with_camera():
"""Test that destroy on an unbuilt scene with cameras doesn't crash."""
scene = gs.Scene(show_viewer=False)
scene.add_entity(morph=gs.morphs.Plane())
scene.add_sensor... | Same. I think it should crash. | else:
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
@pytest.mark.required
def test_destroy_unbuilt_scene_with_camera():
"""Test that destroy on an unbuilt scene with cameras doesn't crash."""
scene = gs.Scene(show_viewer=False)
scene.add_entity(morph=gs.morphs.Plane())
sc... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.143 | bug | 30 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
@pytest.mark.required
def test_destroy_unbuilt_scene_with_camera():
"""T... | Same. I think it should crash. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
"""Test that calling destroy twice on a scene with cameras doesn't crash."""
scene = gs.Scene(show_viewer=False)
scene.add_entity(morph=gs.morphs.Plane())
camera = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(64, 64)))
scene.build()
camera.read()
scene.destroy()
scene.destr... | I don't think this test is relevant but I may be wrong. What are you trying to enforce here? It looks like implementation details. | @pytest.mark.required
def test_destroy_idempotent_with_camera():
"""Test that calling destroy twice on a scene with cameras doesn't crash."""
scene = gs.Scene(show_viewer=False)
camera = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(64, 64)))
scene.build()
camera.read()
scene.destro... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.357 | bug | 130 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: """Test that calling destroy twice on a scene with cameras doesn't crash."""
scene = gs.Scene(show_viewer=False)
scene.add_entity(morp... | I don't think this test is relevant but I may be wrong. What are you trying to enforce here? It looks like implementation details. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
camera.read()
scene.destroy()
scene.destroy() # Should not crash
@pytest.mark.required
def test_cameras_share_metadata():
scene = gs.Scene(show_viewer=False)
scene.add_entity(morph=gs.morphs.Plane())
cam1 = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(64, 64)))
cam2 = scene.... | No local imports except absolutely necessary. | scene.build()
camera.read()
scene.destroy()
# Scene.__del__ calls destroy(), which means it's expected that destroy() will
# be called twice. A crash in destroy() would result in some logspam.
scene.destroy()
@pytest.mark.required
def test_rasterizer_destroy():
scene = gs.Scene(show_viewe... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.071 | suggestion | 45 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: camera.read()
scene.destroy()
scene.destroy() # Should not crash
@pytest.mark.required
def test_cameras_share_metadata():
scen... | No local imports except absolutely necessary. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
scene.build()
cam1.read()
cam2.read()
assert cam1._shared_metadata is cam2._shared_metadata
assert len(cam1._shared_metadata.sensors) == 2
@pytest.mark.required
def test_camera_destroy_cleans_shared_metadata():
import gc
import weakref
scene = gs.Scene(show_viewer=False)
scene.a... | I would rather keep it short to avoid distraction and make it easier to understand what this test is doing at first glance:
```python
offscreen_renderer_ref = weakref.ref(camera._shared_metadata.renderer._renderer)
``` | cam2 = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(32, 32)))
scene.build()
cam1.read()
cam2.read()
offscreen_renderer_ref = weakref.ref(cam1._shared_metadata.renderer._renderer)
scene.destroy()
assert offscreen_renderer_ref() is None
@pytest.mark.required
@pytest.mark.param... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.857 | suggestion | 218 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE:
scene.build()
cam1.read()
cam2.read()
assert cam1._shared_metadata is cam2._shared_metadata
assert len(cam1._shared_metadata... | I would rather keep it short to avoid distraction and make it easier to understand what this test is doing at first glance:
```python
offscreen_renderer_ref = weakref.ref(camera._shared_metadata.renderer._renderer)
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@pytest.mark.required
def test_camera_destroy_cleans_shared_metadata():
import gc
import weakref
scene = gs.Scene(show_viewer=False)
scene.add_entity(morph=gs.morphs.Plane())
camera = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(64, 64)))
scene.build()
camera.read()
share... | Following my previous recommendation, this del is no longer necessary. | assert offscreen_renderer_ref() is None
@pytest.mark.required
@pytest.mark.parametrize("backend", [gs.cuda])
@pytest.mark.skipif(not ENABLE_MADRONA, reason="BatchRenderer is not supported because 'gs_madrona' is not available.")
def test_batch_renderer_destroy():
scene = gs.Scene(show_viewer=False)
# FIXM... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.214 | suggestion | 70 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: @pytest.mark.required
def test_camera_destroy_cleans_shared_metadata():
import gc
import weakref
scene = gs.Scene(show_viewer=False)
... | Following my previous recommendation, this del is no longer necessary. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
def test_camera_destroy_cleans_shared_metadata():
import gc
import weakref
scene = gs.Scene(show_viewer=False)
scene.add_entity(morph=gs.morphs.Plane())
camera = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(64, 64)))
scene.build()
camera.read()
shared_metadata = camera._s... | In my view it is a requirement to release memory right away without waiting for garbage collection. If someone creates another scene right away, it may cause OpenGL related race condition or similar. |
@pytest.mark.required
@pytest.mark.parametrize("backend", [gs.cuda])
@pytest.mark.skipif(not ENABLE_MADRONA, reason="BatchRenderer is not supported because 'gs_madrona' is not available.")
def test_batch_renderer_destroy():
scene = gs.Scene(show_viewer=False)
# FIXME: This test fails without any entities in t... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.571 | bug | 199 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: def test_camera_destroy_cleans_shared_metadata():
import gc
import weakref
scene = gs.Scene(show_viewer=False)
scene.add_entity(m... | In my view it is a requirement to release memory right away without waiting for garbage collection. If someone creates another scene right away, it may cause OpenGL related race condition or similar. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
camera = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(64, 64)))
scene.build()
camera.read()
shared_metadata = camera._shared_metadata
renderer = shared_metadata.renderer
offscreen_renderer = renderer._renderer
offscreen_renderer_ref = weakref.ref(offscreen_renderer)
scene.... | I would rather add a single destroy test that will combine this one and the previous one. | # FIXME: This test fails without any entities in the scene.
scene.add_entity(morph=gs.morphs.Plane())
cam1 = scene.add_sensor(gs.sensors.BatchRendererCameraOptions(res=(64, 64), use_rasterizer=True))
cam2 = scene.add_sensor(gs.sensors.BatchRendererCameraOptions(res=(64, 64), use_rasterizer=True))
s... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.214 | suggestion | 89 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: camera = scene.add_sensor(gs.sensors.RasterizerCameraOptions(res=(64, 64)))
scene.build()
camera.read()
shared_metadata = camera... | I would rather add a single destroy test that will combine this one and the previous one. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
scene.build()
cam1.read()
cam2.read()
shared_metadata = cam1._shared_metadata
offscreen_renderer = shared_metadata.renderer._renderer
offscreen_renderer_ref = weakref.ref(offscreen_renderer)
scene.destroy()
assert shared_metadata.renderer is None
assert shared_metadata.sensors is... | I don't think adding this is relevant. | )
cam1 = scene.add_sensor(gs.sensors.RaytracerCameraOptions(res=(64, 64)))
cam2 = scene.add_sensor(gs.sensors.RaytracerCameraOptions(res=(64, 64)))
scene.build()
cam1.read()
cam2.read()
shared_metadata = cam1._shared_metadata
assert cam1._shared_metadata is cam2._shared_metadata
a... | @@ -317,6 +317,162 @@ def test_batch_renderer(n_envs, png_snapshot):
assert rgb_array_to_png_bytes(data.rgb) == png_snapshot
+@pytest.mark.required
+def test_destroy_unbuilt_scene_with_camera():
+ """Test that destroy on an unbuilt scene with cameras doesn't crash."""
+ scene = gs.Scene(show_view... | tests/test_sensor_camera.py | 26 | Python | 0.071 | suggestion | 38 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE:
scene.build()
cam1.read()
cam2.read()
shared_metadata = cam1._shared_metadata
offscreen_renderer = shared_metadata.renderer.... | I don't think adding this is relevant. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._shared_metadata.context.add_light(light_dict)
camera_wrapper = self._get_camera_wrapper()
self._shared_metadata.renderer.add_camera(camera_wrapper)
self._update_camera_pose()
_B = max(self._manager._sim.n_envs, 1)
w, h = self._options.res
self._sha... | What does this means? I don't think it sound be possible to manually destroy a camera. This should rather be `__del__` and the sensor manager forcibly calling `del camera`. Otherwise the user may assume destroying camera is allowed and part of the public API. You may also add some `_destroy` prefix to make it "private"... |
# ========================== Sensor Lifecycle ==========================
def build(self):
"""Initialize the rasterizer and register this camera."""
super().build()
scene = self._manager._sim.scene
if self._shared_metadata.sensors is None:
self._shared_metadata.sen... | @@ -427,6 +427,23 @@ def build(self):
w, h = self._options.res
self._shared_metadata.image_cache[self._idx] = torch.zeros((_B, h, w, 3), dtype=torch.uint8, device=gs.device)
+ def destroy(self):
+ if not self._shared_metadata.sensors:
+ return
+ self._shared_metadata.sens... | genesis/engine/sensors/camera.py | 26 | Python | 0.643 | question | 335 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: self._shared_metadata.context.add_light(light_dict)
camera_wrapper = self._get_camera_wrapper()
self._shared... | What does this means? I don't think it sound be possible to manually destroy a camera. This should rather be `__del__` and the sensor manager forcibly calling `del camera`. Otherwise the user may assume destroying camera is allowed and part of the public API. You may also add some `_destroy` prefix to make it "private"... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
model=self._options.model,
fov=self._options.fov,
aperture=self._options.aperture,
focus_dist=self._options.focus_dist,
GUI=False,
spp=self._options.spp,
denoise=self._options.denoise,
near=0.05,
far=100.0,
... | I don't think it should work this way. Either all sensors are destroyed or not. They should never be destroyed one by one. Destroy should be a classmethod acting only on shared data basically. | up = transform_by_quat(up, link_quat)
elif self._link is not None:
# Link exists but not built yet - use configured pose as-is (treat as world coordinates for now)
# This will be corrected when move_to_attach is called
pass
self._camera_obj = visualizer.a... | @@ -631,6 +648,12 @@ def build(self):
w, h = self._options.res
self._shared_metadata.image_cache[self._idx] = torch.zeros((_B, h, w, 3), dtype=torch.uint8, device=gs.device)
+ def destroy(self):
+ if not self._shared_metadata.sensors:
+ return
+ self._shared_metadata.sens... | genesis/engine/sensors/camera.py | 26 | Python | 0.429 | suggestion | 192 | 51 | 51 | false | [BUG FIX] Clean up offscreen renderer when scene is destroyed. | 2,282 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | nimrod-gileadi | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Clean up offscreen renderer when scene is destroyed.
BEFORE CODE: model=self._options.model,
fov=self._options.fov,
aperture=self._options.aperture,
focus_dist... | I don't think it should work this way. Either all sensors are destroyed or not. They should never be destroyed one by one. Destroy should be a classmethod acting only on shared data basically. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# Mean density should be in similar range (50-100% of rho0)
assert 0.5 < normalized_mean < 1.0, (
f"Normalized mean density {normalized_mean:.3f} is outside expected range [0.5, 1.0] "
f"for particle_size={particle_size}"
)
# Standard deviation should be bounded
assert normalized_s... | Add a todo to remember that this test should be removed and replace by something that truly validates the physics behind it. |
# Mean density should be in similar range (50-100% of rho0)
assert 0.5 < normalized_mean < 1.0, (
f"Normalized mean density {normalized_mean:.3f} is outside expected range [0.5, 1.0] "
f"for particle_size={particle_size}"
)
# Standard deviation should be bounded
assert normalized_s... | @@ -275,3 +284,29 @@ def test_sph_density_consistency_different_particle_sizes(show_viewer, particle_
assert np.allclose(pressures, 0.0, atol=1e-6), (
f"Initial pressure is not zero for particle_size={particle_size}. Mean: {pressures.mean():.6f}"
)
+
+
+# =============================================... | tests/test_sph.py | 26 | Python | 0.429 | suggestion | 124 | 43 | 44 | false | [BUG FIX] Fix DFSPH solver and add tests. | 2,302 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | erizmr | Review this pull request change.
Language: Python
File: tests/test_sph.py
PR Title: [BUG FIX] Fix DFSPH solver and add tests.
BEFORE CODE:
# Mean density should be in similar range (50-100% of rho0)
assert 0.5 < normalized_mean < 1.0, (
f"Normalized mean density {normalized_mean:.3f} is outside exp... | Add a todo to remember that this test should be removed and replace by something that truly validates the physics behind it. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
@pytest.fixture
def g1_fall(solver, n_envs, gjk):
"""G1 humanoid robot falling from above a plane."""
# This is sufficient, as long as we use sync
duration_warmup = 20.0
duration_record =... | this dt is different from STEP_DT (0.01), so the realtime_factor is wrong |
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
@pytest.fixture
def g1_fall(solver, n_envs, gjk):
"""G1 humanoid robot falling from above a plane."""
import gstaichi as ti
# This is sufficient, as long as we use sync
duration_warmup = ... | @@ -725,6 +724,80 @@ def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
+@pytest.fixture
+def g1_fall(solver, n_envs, gjk):
+ """G1 humanoid robot falling from above a plane."""
+
+ # This is sufficient, as long as we use sync
+ duration_warmup = 20.0
+ dura... | tests/test_rigid_benchmarks.py | 26 | Python | 0.214 | bug | 73 | 51 | 51 | false | [MISC] Add benchmark for Unitree G1. | 2,310 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add benchmark for Unitree G1.
BEFORE CODE:
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
@pytest.fixture
def g1_fall(solver, n_envs, gjk):
... | this dt is different from STEP_DT (0.01), so the realtime_factor is wrong | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
return {"compile_time": compile_time, "runtime_fps": runtime_fps, "realtime_factor": realtime_factor}
@pytest.fixture
def box_pyramid_3(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=3)
@pytest.fixture
def box_pyramid_4(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, g... | solver and gjk are received but unused |
return {"compile_time": compile_time, "runtime_fps": runtime_fps, "realtime_factor": realtime_factor}
@pytest.fixture
def box_pyramid_3(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=3)
@pytest.fixture
def box_pyramid_4(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, g... | @@ -725,6 +724,80 @@ def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
+@pytest.fixture
+def g1_fall(solver, n_envs, gjk): | tests/test_rigid_benchmarks.py | 26 | Python | 0.143 | suggestion | 38 | 51 | 51 | false | [MISC] Add benchmark for Unitree G1. | 2,310 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add benchmark for Unitree G1.
BEFORE CODE:
return {"compile_time": compile_time, "runtime_fps": runtime_fps, "realtime_factor": realtime_factor}
@pytest.fixture
def box_pyramid_3(solver, n_envs, gjk):
... | solver and gjk are received but unused | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@pytest.fixture
def box_pyramid_5(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=5)
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
@pytest.fixture
def g1_fall(solver, n_envs, gjk):
"""G1 humanoid robot falling from abo... | we have get_hf_dataset function, check other tests |
@pytest.fixture
def box_pyramid_5(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=5)
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
@pytest.fixture
def g1_fall(solver, n_envs, gjk):
"""G1 humanoid robot falling from abo... | @@ -725,6 +724,80 @@ def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
+@pytest.fixture
+def g1_fall(solver, n_envs, gjk):
+ """G1 humanoid robot falling from above a plane."""
+
+ # This is sufficient, as long as we use sync
+ duration_warmup = 20.0
+ dura... | tests/test_rigid_benchmarks.py | 26 | Python | 0.214 | suggestion | 50 | 51 | 51 | false | [MISC] Add benchmark for Unitree G1. | 2,310 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add benchmark for Unitree G1.
BEFORE CODE:
@pytest.fixture
def box_pyramid_5(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=5)
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk)... | we have get_hf_dataset function, check other tests | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
return _box_pyramid(solver, n_envs, gjk, n_cubes=5)
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
@pytest.fixture
def g1_fall(solver, n_envs, gjk):
"""G1 humanoid robot falling from above a plane."""
# This is sufficient, as long as we u... | WTF dt 0.005 + iterations=10, completely garbage options. | return _box_pyramid(solver, n_envs, gjk, n_cubes=5)
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
@pytest.fixture
def g1_fall(solver, n_envs, gjk):
"""G1 humanoid robot falling from above a plane."""
import gstaichi as ti
# This is s... | @@ -725,6 +724,80 @@ def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
+@pytest.fixture
+def g1_fall(solver, n_envs, gjk):
+ """G1 humanoid robot falling from above a plane."""
+
+ # This is sufficient, as long as we use sync
+ duration_warmup = 20.0
+ dura... | tests/test_rigid_benchmarks.py | 26 | Python | 0.214 | suggestion | 57 | 51 | 51 | false | [MISC] Add benchmark for Unitree G1. | 2,310 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add benchmark for Unitree G1.
BEFORE CODE: return _box_pyramid(solver, n_envs, gjk, n_cubes=5)
@pytest.fixture
def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)... | WTF dt 0.005 + iterations=10, completely garbage options. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
scene.build(n_envs=n_envs)
compile_time = time.time() - time_start
# Set initial position with robot elevated above ground
init_qpos = torch.zeros((robot.n_qs,), dtype=gs.tc_float, device=gs.device)
init_qpos[2] = 1.0 # z position
init_qpos[3] = 1.0 # quaternion w component
if n_envs > 0:... | Do not use `gs.ti`. This is not supported. Just import ti at the beginning of this "test" function. | scene.build(n_envs=n_envs)
compile_time = time.time() - time_start
# Set initial position with robot elevated above ground
init_qpos = torch.zeros((robot.n_qs,), dtype=gs.tc_float, device=gs.device)
init_qpos[2] = 1.0 # z position
init_qpos[3] = 1.0 # quaternion w component
robot.set_qpos... | @@ -725,6 +724,80 @@ def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
+@pytest.fixture
+def g1_fall(solver, n_envs, gjk):
+ """G1 humanoid robot falling from above a plane."""
+
+ # This is sufficient, as long as we use sync
+ duration_warmup = 20.0
+ dura... | tests/test_rigid_benchmarks.py | 26 | Python | 0.429 | suggestion | 99 | 51 | 51 | false | [MISC] Add benchmark for Unitree G1. | 2,310 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add benchmark for Unitree G1.
BEFORE CODE: scene.build(n_envs=n_envs)
compile_time = time.time() - time_start
# Set initial position with robot elevated above ground
init_qpos = torch.zeros((robo... | Do not use `gs.ti`. This is not supported. Just import ti at the beginning of this "test" function. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
),
show_viewer=False,
show_FPS=False,
)
scene.add_entity(
gs.morphs.Plane(),
)
robot = scene.add_entity(
gs.morphs.MJCF(
**get_file_morph_options(
file=f"{asset_path}/unitree_g1/g1_29dof_rev_1_0.xml",
pos=(0, 0, 1.0),
... | Avoid initializing all envs in the same state, this is not representative of what happens in reality. Either apply random force, different initial velocities, or different initial positions. | ),
show_viewer=False,
show_FPS=False,
)
scene.add_entity(
gs.morphs.Plane(),
)
robot = scene.add_entity(
gs.morphs.MJCF(
**get_file_morph_options(
file=f"{asset_path}/unitree_g1/g1_29dof_rev_1_0.xml",
pos=(0, 0, 1.0),
... | @@ -725,6 +724,80 @@ def box_pyramid_6(solver, n_envs, gjk):
return _box_pyramid(solver, n_envs, gjk, n_cubes=6)
+@pytest.fixture
+def g1_fall(solver, n_envs, gjk):
+ """G1 humanoid robot falling from above a plane."""
+
+ # This is sufficient, as long as we use sync
+ duration_warmup = 20.0
+ dura... | tests/test_rigid_benchmarks.py | 26 | Python | 0.429 | suggestion | 190 | 51 | 51 | false | [MISC] Add benchmark for Unitree G1. | 2,310 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add benchmark for Unitree G1.
BEFORE CODE: ),
show_viewer=False,
show_FPS=False,
)
scene.add_entity(
gs.morphs.Plane(),
)
robot = scene.add_entity(
gs.morp... | Avoid initializing all envs in the same state, this is not representative of what happens in reality. Either apply random force, different initial velocities, or different initial positions. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
import hashlib
import os
import time
from pathlib import Path
from typing import Any
import numpy as np
import pytest
import torch
import wandb
import genesis as gs
import gstaichi as ti
from .utils import (
get_hardware_fingerprint,
get_hf_dataset,
get_platform_fingerprint,
get_git_commit_timestamp,... | Please move taichi import install the fixture. It must not be imported in global scope. | import hashlib
import os
import time
from pathlib import Path
from typing import Any
import numpy as np
import pytest
import torch
import wandb
import genesis as gs
from .utils import (
get_hardware_fingerprint,
get_hf_dataset,
get_platform_fingerprint,
get_git_commit_timestamp,
get_git_commit_in... | @@ -12,9 +10,11 @@
import wandb
import genesis as gs
+import gstaichi as ti | tests/test_rigid_benchmarks.py | 13 | Python | 0.286 | suggestion | 87 | 38 | 38 | false | [MISC] Add benchmark for Unitree G1. | 2,310 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add benchmark for Unitree G1.
BEFORE CODE: import hashlib
import os
import time
from pathlib import Path
from typing import Any
import numpy as np
import pytest
import torch
import wandb
import genesis as gs
im... | Please move taichi import install the fixture. It must not be imported in global scope. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
Whether the mesh is convex.
"""
return self._mesh.is_convex
@property
def metadata(self):
"""
Metadata of the mesh.
"""
return self._metadata
@property
def verts(self):
"""
Vertices of the mesh.
"""
return self._me... | After removing the verts setter, https://github.com/Genesis-Embodied-AI/Genesis/blob/cd84ab234de030ab48f2f30f579f59cc158438d2/genesis/vis/rasterizer_context.py#L562 this line seems to throw the error. | """
return self._mesh.is_convex
@property
def metadata(self):
"""
Metadata of the mesh.
"""
return self._metadata
@property
def verts(self):
"""
Vertices of the mesh.
"""
if self._scale is not None:
return self... | @@ -447,16 +446,10 @@ def verts(self):
"""
Vertices of the mesh.
"""
+ if self._scale is not None:
+ return self._mesh.vertices * self._scale
return self._mesh.vertices
- @verts.setter
- def verts(self, verts):
- """
- Set the vertices of the m... | genesis/engine/mesh.py | 26 | Python | 0.571 | suggestion | 200 | 51 | 51 | false | [BUG FIX] Fix 'Mesh.convert_to_zup' by applying scaling out-of-place. | 2,311 | Genesis-Embodied-AI/Genesis | 26,702 | Python | ACMLCZH | duburcqa | Review this pull request change.
Language: Python
File: genesis/engine/mesh.py
PR Title: [BUG FIX] Fix 'Mesh.convert_to_zup' by applying scaling out-of-place.
BEFORE CODE: Whether the mesh is convex.
"""
return self._mesh.is_convex
@property
def metadata(self):
"""
M... | After removing the verts setter, https://github.com/Genesis-Embodied-AI/Genesis/blob/cd84ab234de030ab48f2f30f579f59cc158438d2/genesis/vis/rasterizer_context.py#L562 this line seems to throw the error. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# Overwrite surface color by original color specified in URDF file only if necessary
if geom_is_col:
geom_surface = gs.surfaces.Collision()
elif (
not geom_is_col
and getattr(geom_prop, "material") is not None
... | we assume non-glb mesh will never have `file_meshes_are_zup=False`? Maybe it's okay | # Overwrite surface color by original color specified in URDF file only if necessary
if geom_is_col:
geom_surface = gs.surfaces.Collision()
elif (
not geom_is_col
and getattr(geom_prop, "material") is not None
... | @@ -128,26 +130,20 @@ def parse_urdf(morph, surface):
else:
geom_surface = surface
- mesh_path = urdfpy.utils.get_filename(os.path.dirname(path), geometry.filename)
+ mesh_path = urdfpy.utils.get_filename(parent_dir, geometry.filename)
... | genesis/utils/urdf.py | 26 | Python | 0.357 | suggestion | 83 | 51 | 51 | false | [BUG FIX] Revert out-of-place mesh scaling and implement a new fix. | 2,316 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | duburcqa | Review this pull request change.
Language: Python
File: genesis/utils/urdf.py
PR Title: [BUG FIX] Revert out-of-place mesh scaling and implement a new fix.
BEFORE CODE: # Overwrite surface color by original color specified in URDF file only if necessary
if geom_is_col:
... | we assume non-glb mesh will never have `file_meshes_are_zup=False`? Maybe it's okay | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
if cfg.get("quality", True):
r = cfg.get("minratio", 1.1)
di = cfg.get("mindihedral", 10)
flags.append(f"q{r}/{di}")
a = cfg.get("maxvolume", -1.0)
if a > 0:
flags.append(f"a{a}")
o = cfg.get("order", 1)
if o != 1:
flags.append(f"o{o}")
if cfg.get("nob... | Note: This can be done more efficiently by directly passing faces via:
```pv
points = mesh.vertices
faces = mesh._connectivity_array.reshape(-1, 3).astype(np.int32, copy=False)
tet = tetgen.TetGen(points, faces)
``` | if cfg.get("quality", True):
r = cfg.get("minratio", 1.1)
di = cfg.get("mindihedral", 10)
flags.append(f"q{r}/{di}")
a = cfg.get("maxvolume", -1.0)
if a > 0:
flags.append(f"a{a}")
o = cfg.get("order", 1)
if o != 1:
flags.append(f"o{o}")
if cfg.get("nobi... | @@ -992,9 +991,6 @@ def make_tetgen_switches(cfg):
def tetrahedralize_mesh(mesh, tet_cfg):
- if platform.machine() == "aarch64":
- gs.raise_exception("This method is not support on Linux ARM because 'tetgen' module is crashing.")
-
# Importing pyvista and tetgen are very slow to import and not used ... | genesis/utils/mesh.py | 26 | Python | 0.929 | performance | 220 | 51 | 51 | false | [MISC] Add support of tetgen on Linux ARM. | 2,317 | Genesis-Embodied-AI/Genesis | 26,702 | Python | akaszynski | duburcqa | Review this pull request change.
Language: Python
File: genesis/utils/mesh.py
PR Title: [MISC] Add support of tetgen on Linux ARM.
BEFORE CODE:
if cfg.get("quality", True):
r = cfg.get("minratio", 1.1)
di = cfg.get("mindihedral", 10)
flags.append(f"q{r}/{di}")
a = cfg.get("maxvolum... | Note: This can be done more efficiently by directly passing faces via:
```pv
points = mesh.vertices
faces = mesh._connectivity_array.reshape(-1, 3).astype(np.int32, copy=False)
tet = tetgen.TetGen(points, faces)
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
Raises
------
Exception
If the morph type is unsupported.
"""
from genesis.engine.materials.FEM.cloth import Cloth as ClothMaterial
is_cloth = isinstance(self.material, ClothMaterial)
self._uvs = None
if is_cloth:
# Cloth: load s... | I would recommend:
```python
self._uvs = np.array(mesh.visual.uv, dtype=gs.np_float, copy=True)
```
or just:
```python
self._uvs = mesh.visual.uv.astype(gs.np_float, copy=True)
```
More efficient as it avoids allocating memory twice.
|
Raises
------
Exception
If the morph type is unsupported.
"""
from genesis.engine.materials.FEM.cloth import Cloth as ClothMaterial
is_cloth = isinstance(self.material, ClothMaterial)
self._uvs = None
if is_cloth:
# Cloth: load s... | @@ -400,6 +401,13 @@ def sample(self):
faces = mesh.faces
# For cloth, we store faces as "elements" (treating them as surface elements)
self.instantiate(verts, faces)
+
+ # Load UVs from mesh (1:1 mapping for cloth)
+ try:
+ ... | genesis/engine/entities/fem_entity.py | 26 | Python | 1 | performance | 245 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE:
Raises
------
Exception
If the morph type is unsupported.
"""
fr... | I would recommend:
```python
self._uvs = np.array(mesh.visual.uv, dtype=gs.np_float, copy=True)
```
or just:
```python
self._uvs = mesh.visual.uv.astype(gs.np_float, copy=True)
```
More efficient as it avoids allocating memory twice.
| You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
------
Exception
If the morph type is unsupported.
"""
from genesis.engine.materials.FEM.cloth import Cloth as ClothMaterial
is_cloth = isinstance(self.material, ClothMaterial)
self._uvs = None
if is_cloth:
# Cloth: load surface mesh dire... | Could you add a comment to explain which attribute may not exist exactly, in which scenario it would be the case (and why this scenario must be supported if not obvious), and why just ignoring UV is the right approach if not obvious. | ------
Exception
If the morph type is unsupported.
"""
from genesis.engine.materials.FEM.cloth import Cloth as ClothMaterial
is_cloth = isinstance(self.material, ClothMaterial)
self._uvs = None
if is_cloth:
# Cloth: load surface mesh dire... | @@ -400,6 +401,13 @@ def sample(self):
faces = mesh.faces
# For cloth, we store faces as "elements" (treating them as surface elements)
self.instantiate(verts, faces)
+
+ # Load UVs from mesh (1:1 mapping for cloth)
+ try:
+ ... | genesis/engine/entities/fem_entity.py | 26 | Python | 0.5 | suggestion | 233 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: ------
Exception
If the morph type is unsupported.
"""
from genesis.engin... | Could you add a comment to explain which attribute may not exist exactly, in which scenario it would be the case (and why this scenario must be supported if not obvious), and why just ignoring UV is the right approach if not obvious. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._uvs = None
else:
gs.raise_exception(f"Cloth material only supports Mesh morph. Got: {self.morph}.")
else:
# Regular FEM: tetrahedralize mesh
if isinstance(self.morph, gs.options.morphs.Sphere):
verts, elems = eu.sphere... | Why can you assign uvs directly while verts, elems must go through further processing? | if isinstance(mesh.visual, trimesh.visual.texture.TextureVisuals) and mesh.visual.uv is not None:
self._uvs = mesh.visual.uv.astype(gs.np_float, copy=False)
else:
gs.raise_exception(f"Cloth material only supports Mesh morph. Got: {self.morph}.")
el... | @@ -419,11 +427,12 @@ def sample(self):
elif isinstance(self.morph, gs.options.morphs.Cylinder):
verts, elems = eu.cylinder_to_elements()
elif isinstance(self.morph, gs.options.morphs.Mesh):
- verts, elems = eu.mesh_to_elements(
+ verts, elems, se... | genesis/engine/entities/fem_entity.py | 26 | Python | 0.214 | question | 86 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: self._uvs = None
else:
gs.raise_exception(f"Cloth material only suppo... | Why can you assign uvs directly while verts, elems must go through further processing? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._uvs = None
else:
gs.raise_exception(f"Cloth material only supports Mesh morph. Got: {self.morph}.")
else:
# Regular FEM: tetrahedralize mesh
if isinstance(self.morph, gs.options.morphs.Sphere):
verts, elems = eu.sphere... | Just return UVS systematically and update all places where this function is being called. Use `, _ = [...]` to assign to new unused return value. | if isinstance(mesh.visual, trimesh.visual.texture.TextureVisuals) and mesh.visual.uv is not None:
self._uvs = mesh.visual.uv.astype(gs.np_float, copy=False)
else:
gs.raise_exception(f"Cloth material only supports Mesh morph. Got: {self.morph}.")
el... | @@ -419,11 +427,12 @@ def sample(self):
elif isinstance(self.morph, gs.options.morphs.Cylinder):
verts, elems = eu.cylinder_to_elements()
elif isinstance(self.morph, gs.options.morphs.Mesh):
- verts, elems = eu.mesh_to_elements(
+ verts, elems, se... | genesis/engine/entities/fem_entity.py | 26 | Python | 0.571 | suggestion | 145 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: self._uvs = None
else:
gs.raise_exception(f"Cloth material only suppo... | Just return UVS systematically and update all places where this function is being called. Use `, _ = [...]` to assign to new unused return value. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
v_start=self._v_start,
el_start=self._el_start,
s_start=self._s_start,
verts=verts_numpy,
elems=elems_np,
tri2v=self._surface_tri_np,
tri2el=self._surface_el_np,
)
# Add UVs to solver if ... | Just extend `_add_to_solver` with the content of `_add_uvs_to_solver`. | mat_idx=self._material.idx,
mat_mu=self._material.mu,
mat_lam=self._material.lam,
mat_rho=self._material.rho,
mat_friction_mu=self._material.friction_mu,
n_surfaces=self._n_surfaces,
v_start=self._v_start,
... | @@ -477,8 +486,23 @@ def _add_to_solver(self, in_backward=False):
tri2el=self._surface_el_np,
)
+ # Add UVs to solver if available
+ self._add_uvs_to_solver()
+
self.active = True
+ def _add_uvs_to_solver(self):
+ """
+ Copy UV coordinates to the... | genesis/engine/entities/fem_entity.py | 26 | Python | 0.357 | suggestion | 71 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: v_start=self._v_start,
el_start=self._el_start,
s_start=self._s_start... | Just extend `_add_to_solver` with the content of `_add_uvs_to_solver`. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
mat_rho=self._material.rho,
mat_friction_mu=self._material.friction_mu,
n_surfaces=self._n_surfaces,
v_start=self._v_start,
el_start=self._el_start,
s_start=self._s_start,
verts=verts_numpy,
e... | This should already be enforced by `instantiate` | elems_np = self.elems.astype(gs.np_int, copy=False)
self._solver._kernel_add_elements(
f=self._sim.cur_substep_local,
mat_idx=self._material.idx,
mat_mu=self._material.mu,
mat_lam=self._material.lam,
mat_rho=self._ma... | @@ -477,8 +486,23 @@ def _add_to_solver(self, in_backward=False):
tri2el=self._surface_el_np,
)
+ # Add UVs to solver if available
+ self._add_uvs_to_solver()
+
self.active = True
+ def _add_uvs_to_solver(self):
+ """
+ Copy UV coordinates to the... | genesis/engine/entities/fem_entity.py | 26 | Python | 0.286 | suggestion | 49 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: mat_rho=self._material.rho,
mat_friction_mu=self._material.friction_mu,
... | This should already be enforced by `instantiate` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
n_surfaces=self._n_surfaces,
v_start=self._v_start,
el_start=self._el_start,
s_start=self._s_start,
verts=verts_numpy,
elems=elems_np,
tri2v=self._surface_tri_np,
tri2el=self._surface_el_np,
... | This is extremely inefficient (one kernel call per assignment). You should modify `_kernel_add_elements` to take into account `uv`. Just pass an 0-length vector if unspecified and check this condition inside the kernel to avoid out of bound access. | f=self._sim.cur_substep_local,
mat_idx=self._material.idx,
mat_mu=self._material.mu,
mat_lam=self._material.lam,
mat_rho=self._material.rho,
mat_friction_mu=self._material.friction_mu,
n_surfaces=self._n_surf... | @@ -477,8 +486,23 @@ def _add_to_solver(self, in_backward=False):
tri2el=self._surface_el_np,
)
+ # Add UVs to solver if available
+ self._add_uvs_to_solver()
+
self.active = True
+ def _add_uvs_to_solver(self):
+ """
+ Copy UV coordinates to the... | genesis/engine/entities/fem_entity.py | 26 | Python | 0.643 | performance | 249 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: n_surfaces=self._n_surfaces,
v_start=self._v_start,
el_start=self._el... | This is extremely inefficient (one kernel call per assignment). You should modify `_kernel_add_elements` to take into account `uv`. Just pass an 0-length vector if unspecified and check this condition inside the kernel to avoid out of bound access. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
import gstaichi as ti
import numpy as np
import trimesh
import genesis as gs
import genesis.utils.geom as gu
import genesis.utils.mesh as mu
from genesis.engine.entities.particle_entity import ParticleEntity
class PBDBaseEntity(ParticleEntity):
"""
Base class for PBD entity.
"""
def _add_to_solver(s... | I have more or less the same remarks than for FEM. | import gstaichi as ti
import numpy as np
import trimesh
import genesis as gs
import genesis.utils.geom as gu
import genesis.utils.mesh as mu
from genesis.engine.entities.particle_entity import ParticleEntity
class PBDBaseEntity(ParticleEntity):
"""
Base class for PBD entity.
"""
def _add_to_solver(s... | genesis/engine/entities/pbd_entity.py | 1 | Python | 0.214 | suggestion | 50 | 26 | 26 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/pbd_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: import gstaichi as ti
import numpy as np
import trimesh
import genesis as gs
import genesis.utils.geom as gu
import ... | I have more or less the same remarks than for FEM. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
0,
self._ckpt[ckpt_name]["pos"],
self._ckpt[ckpt_name]["vel"],
self._ckpt[ckpt_name]["active"],
)
for entity in self._entities:
entity.load_ckpt(ckpt_name=ckpt_name)
# ------------------------------------------... | Just enforce return uvs systematically now. This is private API. No backward compatibility concerns. | )
for entity in self._entities:
entity.load_ckpt(ckpt_name=ckpt_name)
# ------------------------------------------------------------------------------------
# --------------------------------------- io -----------------------------------------
# --------------------... | @@ -1092,11 +1099,14 @@ def get_state(self, f):
state = None
return state
- def get_state_render(self, f):
+ def get_state_render(self, f, include_uvs=False): | genesis/engine/solvers/fem_solver.py | 26 | Python | 0.357 | suggestion | 100 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/solvers/fem_solver.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: 0,
self._ckpt[ckpt_name]["pos"],
self._ckpt[ckpt_name]["vel"],
... | Just enforce return uvs systematically now. This is private API. No backward compatibility concerns. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
def get_hashkey(*args):
hasher = hashlib.sha256()
for arg in (*args, gs.__version__.encode()):
if isinstance(arg, Path):
file_stats = arg.stat()
arg = (str(arg).encode(), file_stats.st_size, file_stats.st_mtime)
if isinstance(arg, str):
arg = arg.encode()
... | Why skipping texture would make sense in the first place. This is weird. I would suggest to just load texture systematically, and if it fails, fallback to loading without texture, with a `gs.logger.warning` stating that texture will be skipped because they failed to load. | def get_hashkey(*args):
hasher = hashlib.sha256()
for arg in (*args, gs.__version__.encode()):
if isinstance(arg, Path):
file_stats = arg.stat()
arg = (str(arg).encode(), file_stats.st_size, file_stats.st_mtime)
if isinstance(arg, str):
arg = arg.encode()
... | @@ -180,9 +180,9 @@ def get_hashkey(*args):
return hasher.hexdigest()
-def load_mesh(file):
+def load_mesh(file, skip_texture=True):
if isinstance(file, (str, Path)):
- return trimesh.load(file, force="mesh", skip_texture=True)
+ return trimesh.load(file, force="mesh", skip_texture=skip_text... | genesis/utils/mesh.py | 26 | Python | 0.643 | bug | 272 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/utils/mesh.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: def get_hashkey(*args):
hasher = hashlib.sha256()
for arg in (*args, gs.__version__.encode()):
if isinstance(arg, Pat... | Why skipping texture would make sense in the first place. This is weird. I would suggest to just load texture systematically, and if it fails, fallback to loading without texture, with a `gs.logger.warning` stating that texture will be skipped because they failed to load. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
If the morph type is unsupported.
"""
from genesis.engine.materials.FEM.cloth import Cloth as ClothMaterial
is_cloth = isinstance(self.material, ClothMaterial)
self._uvs = None
if is_cloth:
# Cloth: load surface mesh directly (no tetrahedralization)
... | * Everything must fit into 120 chars limit, including comments
* We prefer explicit check instead of try-except in this context
```python
# Load UVs from mesh (1:1 mapping for cloth).
# UVs are not always available in 3D file, in case they are missing we set the entity UVs to None when UVs are None,
# the solver will ... | If the morph type is unsupported.
"""
from genesis.engine.materials.FEM.cloth import Cloth as ClothMaterial
is_cloth = isinstance(self.material, ClothMaterial)
self._uvs = None
if is_cloth:
# Cloth: load surface mesh directly (no tetrahedralization)
... | @@ -400,6 +401,15 @@ def sample(self):
faces = mesh.faces
# For cloth, we store faces as "elements" (treating them as surface elements)
self.instantiate(verts, faces)
+
+ # Load UVs from mesh (1:1 mapping for cloth)
+ try:
+ ... | genesis/engine/entities/fem_entity.py | 26 | Python | 1 | suggestion | 601 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: If the morph type is unsupported.
"""
from genesis.engine.materials.FEM.cloth import Clot... | * Everything must fit into 120 chars limit, including comments
* We prefer explicit check instead of try-except in this context
```python
# Load UVs from mesh (1:1 mapping for cloth).
# UVs are not always available in 3D file, in case they are missing we set the entity UVs to None when UVs are None,
# the solver will ... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# We can't generate UVs for newly created internal vertices as it doesn't make sense but they're
# not used for rendering so it's fine.
verts, elems, self._uvs = eu.mesh_to_elements(
file=self._morph.file,
pos=self._morph.pos,
... | Why are you not just setting `self._uvs = np.zeros((0, 2), dtype=gs.np_float)` instead of `None` in the first place? Then you could pass `self._uvs` directly instead of going through `uvs_np`.
Because you want the property `.uvs` to return None if not specified? If so, I'm ok, I don't have a strong opinion. | # not used for rendering so it's fine.
verts, elems, self._uvs = eu.mesh_to_elements(
file=self._morph.file,
pos=self._morph.pos,
scale=self._morph.scale,
tet_cfg=self.tet_cfg,
)
e... | @@ -443,6 +457,7 @@ def _add_to_solver(self, in_backward=False):
# Convert to appropriate numpy array types
verts_numpy = tensor_to_array(self.init_positions, dtype=gs.np_float)
+ uvs_np = self._uvs if self._uvs is not None else np.zeros((0, 2), dtype=gs.np_float) | genesis/engine/entities/fem_entity.py | 26 | Python | 0.786 | question | 310 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/fem_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: # We can't generate UVs for newly created internal vertices as it doesn't make sense but they're
... | Why are you not just setting `self._uvs = np.zeros((0, 2), dtype=gs.np_float)` instead of `None` in the first place? Then you could pass `self._uvs` directly instead of going through `uvs_np`.
Because you want the property `.uvs` to return None if not specified? If so, I'm ok, I don't have a strong opinion. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
import genesis.utils.geom as gu
import genesis.utils.mesh as mu
from genesis.engine.entities.particle_entity import ParticleEntity
class PBDBaseEntity(ParticleEntity):
"""
Base class for PBD entity.
"""
def _add_to_solver(self):
super()._add_to_solver()
self._add_uvs_and_faces_to_solv... | Single line if it fits: `self._kernel_add_uvs_and_faces_to_solver(uvs=uvs, faces=self._vfaces)` | import genesis.utils.geom as gu
import genesis.utils.mesh as mu
from genesis.engine.entities.particle_entity import ParticleEntity
class PBDBaseEntity(ParticleEntity):
"""
Base class for PBD entity.
"""
def _add_to_solver(self):
super()._add_to_solver()
# Get UVs from vmesh (may be N... | @@ -13,6 +13,46 @@ class PBDBaseEntity(ParticleEntity):
Base class for PBD entity.
"""
+ def _add_to_solver(self):
+ super()._add_to_solver()
+ self._add_uvs_and_faces_to_solver()
+
+ def _add_uvs_and_faces_to_solver(self):
+ """Add UV coordinates and face indices to the solver's ... | genesis/engine/entities/pbd_entity.py | 26 | Python | 0.357 | suggestion | 95 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/pbd_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: import genesis.utils.geom as gu
import genesis.utils.mesh as mu
from genesis.engine.entities.particle_entity import P... | Single line if it fits: `self._kernel_add_uvs_and_faces_to_solver(uvs=uvs, faces=self._vfaces)` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._kernel_add_uvs_and_faces_to_solver(
uvs=uvs,
faces=self._vfaces,
)
@ti.kernel
def _kernel_add_uvs_and_faces_to_solver(
self,
uvs: ti.types.ndarray(element_dim=1),
faces: ti.types.ndarray(element_dim=1),
):
# Copy UVs to solver's... | Remove cast, it should not be necessary. You should rather cast the input argument before calling this kernel. | @ti.kernel
def _kernel_add_uvs_and_faces_to_solver(
self,
uvs: ti.types.ndarray(element_dim=1),
faces: ti.types.ndarray(element_dim=1),
):
# Copy UVs to solver's global UV buffer (skip if no UVs provided)
n_uvs = uvs.shape[0]
for i_vv_ in range(n_uvs):
... | @@ -13,6 +13,46 @@ class PBDBaseEntity(ParticleEntity):
Base class for PBD entity.
"""
+ def _add_to_solver(self):
+ super()._add_to_solver()
+ self._add_uvs_and_faces_to_solver()
+
+ def _add_uvs_and_faces_to_solver(self):
+ """Add UV coordinates and face indices to the solver's ... | genesis/engine/entities/pbd_entity.py | 26 | Python | 0.571 | suggestion | 110 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/engine/entities/pbd_entity.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE:
self._kernel_add_uvs_and_faces_to_solver(
uvs=uvs,
faces=self._vfaces,
)
... | Remove cast, it should not be necessary. You should rather cast the input argument before calling this kernel. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
gs.logger.info("Ignoring corrupted cache.")
if not is_cached_loaded:
with gs.logger.timer(f"Tetrahedralization with configuration {tet_cfg} and generating `.tet` file:"):
verts, elems = mu.tetrahedralize_mesh(mesh, tet_cfg)
os.makedirs(os.path.dirname(tet_file_path), ex... | `uvs = np.pad(original_uvs, ((0, len(verts) - n_original), (0, 0)))` | gs.logger.info("Ignoring corrupted cache.")
if not is_cached_loaded:
with gs.logger.timer(f"Tetrahedralization with configuration {tet_cfg} and generating `.tet` file:"):
verts, elems = mu.tetrahedralize_mesh(mesh, tet_cfg)
os.makedirs(os.path.dirname(tet_file_path), ex... | @@ -54,12 +55,27 @@ def mesh_to_elements(file, pos=(0, 0, 0), scale=1.0, tet_cfg=dict()):
verts, elems = mu.tetrahedralize_mesh(mesh, tet_cfg)
os.makedirs(os.path.dirname(tet_file_path), exist_ok=True)
- with open(tet_file_path, "wb") as file:
- pkl.dump((verts, ele... | genesis/utils/element.py | 26 | Python | 0.357 | suggestion | 68 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: genesis/utils/element.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: gs.logger.info("Ignoring corrupted cache.")
if not is_cached_loaded:
with gs.logger.timer(f"Tetrahedraliz... | `uvs = np.pad(original_uvs, ((0, len(verts) - n_original), (0, 0)))` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
substeps=6,
),
pbd_options=gs.options.PBDOptions(
particle_size=1e-2,
),
fem_options=gs.options.FEMOptions(
use_implicit_solver=True, # Implicit solver allows for larger timestep without failure on GPU backend
n_pcg_iterations=40, # Reduc... | Use `get_hf_dataset`, eg:
```python
asset_path = get_hf_dataset(pattern="heavy_three_joint_link.xml")
robot_1 = scene.add_entity(
gs.morphs.MJCF(
file=f"{asset_path}/heavy_three_joint_link.xml",
pos=(-0.2, -0.26, 0.0),
scale=0.3,
),
)
``` | substeps=6,
),
pbd_options=gs.options.PBDOptions(
particle_size=1e-2,
),
fem_options=gs.options.FEMOptions(
use_implicit_solver=True, # Implicit solver allows for larger timestep without failure on GPU backend
n_pcg_iterations=40, # Reduc... | @@ -1543,3 +1543,90 @@ def test_add_camera_vs_interactive_viewer_consistency(add_box, renderer_type, sh
f"interactive viewer brightness ({viewer_brightness:.2f}), "
f"but ratio is {brightness_ratio:.2f}"
)
+
+
+@pytest.mark.parametrize("renderer_type", [RENDERER_TYPE.RASTERIZER, RENDERER_TYPE.RAY... | tests/test_render.py | 26 | Python | 0.857 | suggestion | 302 | 51 | 51 | false | [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities. | 2,323 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alelievr | Review this pull request change.
Language: Python
File: tests/test_render.py
PR Title: [FEATURE] Track mesh UVs in FEM/PBD solvers for rendering of deformable entities.
BEFORE CODE: substeps=6,
),
pbd_options=gs.options.PBDOptions(
particle_size=1e-2,
),
fem_o... | Use `get_hf_dataset`, eg:
```python
asset_path = get_hf_dataset(pattern="heavy_three_joint_link.xml")
robot_1 = scene.add_entity(
gs.morphs.MJCF(
file=f"{asset_path}/heavy_three_joint_link.xml",
pos=(-0.2, -0.26, 0.0),
scale=0.3,
),
)
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
scene.rigid_solver._func_forward_kinematics_entity(
i_e=robot.idx, envs_idx=torch.tensor((1,), dtype=gs.tc_int, device=gs.device)
)
assert_allclose(index_finger_distal.get_pos(envs_idx=(1,)), index_finger_pos, tol=TOL)
assert_allclose(middle_finger_distal.get_pos(envs_idx=(1,)), middle_finger_po... | can we add a variant for (n_envs > 1) | envs_idx=(1,),
pos_tol=TOL,
rot_tol=TOL,
return_error=True,
)
assert qpos.shape == (1, robot.n_qs)
assert err.shape == (1, 3, 6)
assert err.abs().max() < TOL
if show_viewer:
robot.set_qpos(qpos, envs_idx=(1,))
scene.visualizer.update()
links_pos, ... | @@ -1575,6 +1575,127 @@ def test_multilink_inverse_kinematics(show_viewer):
assert_allclose(wrist.get_pos(envs_idx=(1,)), wrist_pos, tol=TOL)
+@pytest.mark.required
+@pytest.mark.parametrize("backend", [gs.cpu, gs.gpu])
+def test_inverse_kinematics_local_point(show_viewer):
+ """Test IK with local_point par... | tests/test_rigid_physics.py | 26 | Python | 0.071 | suggestion | 37 | 51 | 51 | false | [FEATURE] #2332 Implement Local Offset in Inverse Kinematics | 2,333 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | alexis779 | Review this pull request change.
Language: Python
File: tests/test_rigid_physics.py
PR Title: [FEATURE] #2332 Implement Local Offset in Inverse Kinematics
BEFORE CODE: scene.rigid_solver._func_forward_kinematics_entity(
i_e=robot.idx, envs_idx=torch.tensor((1,), dtype=gs.tc_int, device=gs.device)
)
... | can we add a variant for (n_envs > 1) | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
gs.raise_exception("Target link not provided.")
poss = list(poss) if poss is not None else []
if not poss:
poss = [None for _ in range(n_links)]
pos_mask = [False, False, False]
elif len(poss) != n_links:
gs.raise_exception("Accepting only `poss` ... | `local_points[i] = torch.as_tensor(lp, dtype=gs.tc_float, device=gs.device)`
`contiguous()` is useless here, you are stacking afterward anyway. | gs.raise_exception("Target link not provided.")
poss = list(poss) if poss is not None else []
if not poss:
poss = [None for _ in range(n_links)]
pos_mask = [False, False, False]
elif len(poss) != n_links:
gs.raise_exception("Accepting only `poss` ... | @@ -1611,6 +1623,18 @@ def inverse_kinematics_multilink(
elif len(quats) != n_links:
gs.raise_exception("Accepting only `quats` with length equal to `links` or empty list.")
+ # Process local_points - default to origin [0, 0, 0] for each link
+ local_points = list(local_points) if ... | genesis/engine/entities/rigid_entity/rigid_entity.py | 26 | Python | 0.714 | suggestion | 144 | 51 | 51 | false | [FEATURE] #2332 Implement Local Offset in Inverse Kinematics | 2,333 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alexis779 | Review this pull request change.
Language: Python
File: genesis/engine/entities/rigid_entity/rigid_entity.py
PR Title: [FEATURE] #2332 Implement Local Offset in Inverse Kinematics
BEFORE CODE: gs.raise_exception("Target link not provided.")
poss = list(poss) if poss is not None else []
... | `local_points[i] = torch.as_tensor(lp, dtype=gs.tc_float, device=gs.device)`
`contiguous()` is useless here, you are stacking afterward anyway. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
return_error=True,
)
assert qpos.shape == (1, robot.n_qs)
assert err.shape == (1, 3, 6)
assert err.abs().max() < TOL
if show_viewer:
robot.set_qpos(qpos, envs_idx=(1,))
scene.visualizer.update()
links_pos, links_quat = robot.forward_kinematics(qpos, envs_idx=(1,))
as... | Remove this. I don’t think we need to test this on both backends systematically. Then we merge. | return_error=True,
)
assert qpos.shape == (1, robot.n_qs)
assert err.shape == (1, 3, 6)
assert err.abs().max() < TOL
if show_viewer:
robot.set_qpos(qpos, envs_idx=(1,))
scene.visualizer.update()
links_pos, links_quat = robot.forward_kinematics(qpos, envs_idx=(1,))
as... | @@ -1593,6 +1593,150 @@ def test_multilink_inverse_kinematics(show_viewer):
assert_allclose(wrist.get_pos(envs_idx=(1,)), wrist_pos, tol=TOL)
+@pytest.mark.required
+@pytest.mark.parametrize("n_envs", [0, 2])
+@pytest.mark.parametrize("backend", [gs.cpu, gs.gpu]) | tests/test_rigid_physics.py | 26 | Python | 0.286 | refactor | 95 | 51 | 51 | false | [FEATURE] #2332 Implement Local Offset in Inverse Kinematics | 2,333 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alexis779 | Review this pull request change.
Language: Python
File: tests/test_rigid_physics.py
PR Title: [FEATURE] #2332 Implement Local Offset in Inverse Kinematics
BEFORE CODE: return_error=True,
)
assert qpos.shape == (1, robot.n_qs)
assert err.shape == (1, 3, 6)
assert err.abs().max() < TOL
if ... | Remove this. I don’t think we need to test this on both backends systematically. Then we merge. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@dataclass
class RasterizerCameraSharedMetadata(RigidSensorMetadataMixin, SharedSensorMetadata):
"""Shared metadata for all Rasterizer cameras."""
# Rasterizer instance
renderer: Optional["Rasterizer"] = None
# RasterizerContext instance
context: Optional["RasterizerContext"] = None
# List of l... | This change is useless. Revert it. |
@dataclass
class RasterizerCameraSharedMetadata(RigidSensorMetadataMixin, SharedSensorMetadata):
"""Shared metadata for all Rasterizer cameras."""
# Rasterizer instance
renderer: Optional["Rasterizer"] = None
# RasterizerContext instance
context: Optional["RasterizerContext"] = None
# List of ... | @@ -160,10 +160,10 @@ def destroy(self):
if self.renderer is not None:
self.renderer.destroy()
- self.renderer = None
if self.context is not None:
self.context.destroy()
- self.context = None
+ self.renderer = None
+ self.context = None | genesis/engine/sensors/camera.py | 26 | Python | 0.143 | suggestion | 34 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: @dataclass
class RasterizerCameraSharedMetadata(RigidSensorMetadataMixin, SharedSensorMetadata):
... | This change is useless. Revert it. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# ========================== Sensor Lifecycle ==========================
def build(self):
"""Initialize the rasterizer and register this camera."""
super().build()
scene = self._manager._sim.scene
if self._shared_metadata.sensors is None:
self._shared_metadata.sens... | We should support multiple OpenGL context in parallel. Here you are hiding a bug while complexifying the implementation. |
# ========================== Sensor Lifecycle ==========================
def build(self):
"""Initialize the rasterizer and register this camera."""
super().build()
scene = self._manager._sim.scene
if self._shared_metadata.sensors is None:
self._shared_metadata.sen... | @@ -431,29 +432,47 @@ def build(self):
self._shared_metadata.lights = gs.List()
self._shared_metadata.image_cache = {}
- # Create standalone rasterizer
- self._shared_metadata.context = self._create_standalone_context(scene)
-
- self._shared_metadata.renderer... | genesis/engine/sensors/camera.py | 26 | Python | 0.571 | bug | 120 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: # ========================== Sensor Lifecycle ==========================
def build(self):... | We should support multiple OpenGL context in parallel. Here you are hiding a bug while complexifying the implementation. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
morph = gs.morphs.Sphere(pos=pos, radius=radius)
scene.add_mesh_light(
morph=morph,
color=(*color, 1.0),
intensity=intensity,
revert_dir=revert_dir,
double_sided=double_sided,
cutoff=cutoff,
)
def _render_current_state... | Why are you forcing a copy. This is wrong. | def _render_current_state(self):
"""Perform the actual render for the current state."""
if self._link is not None:
self._camera_obj.move_to_attach()
rgb_arr, _, _, _ = self._camera_obj.render(
rgb=True,
depth=False,
segmentation=False,
... | @@ -699,7 +727,7 @@ def _render_current_state(self):
antialiasing=False,
force_render=True,
)
- rgb_tensor = torch.from_numpy(rgb_arr).to(dtype=torch.uint8, device=gs.device)
+ rgb_tensor = torch.from_numpy(rgb_arr.copy()).to(dtype=torch.uint8, device=gs.device) | genesis/engine/sensors/camera.py | 26 | Python | 0.071 | bug | 42 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE:
morph = gs.morphs.Sphere(pos=pos, radius=radius)
scene.add_mesh_light(
... | Why are you forcing a copy. This is wrong. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
assert shared_metadata.renderer is None
@pytest.mark.required
@pytest.mark.parametrize("backend", [gs.cuda])
@pytest.mark.skipif(not ENABLE_RAYTRACER, reason="RayTracer is not supported because 'LuisaRenderPy' is not available.")
def test_raytracer_attached_without_offset_T():
"""Test that RaytracerCameraSens... | I would rather update `test_raytracer` to add `pos=(0.0, 0.0, 2.0),` to `camera_1` instead of adding a new test? Because every single test we add is very costly (it requires building an entire scene), so it can be avoided, it is much better. | assert shared_metadata.renderer is None
@pytest.mark.required
@pytest.mark.parametrize("backend", [gs.cuda])
@pytest.mark.skipif(not ENABLE_RAYTRACER, reason="RayTracer is not supported because 'LuisaRenderPy' is not available.")
def test_raytracer_attached_without_offset_T():
"""Test that RaytracerCameraSens... | @@ -418,6 +418,31 @@ def test_raytracer_destroy():
assert shared_metadata.renderer is None
+@pytest.mark.required
+@pytest.mark.parametrize("backend", [gs.cuda])
+@pytest.mark.skipif(not ENABLE_RAYTRACER, reason="RayTracer is not supported because 'LuisaRenderPy' is not available.")
+def test_raytracer_attache... | tests/test_sensor_camera.py | 26 | Python | 0.643 | suggestion | 241 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: assert shared_metadata.renderer is None
@pytest.mark.required
@pytest.mark.parametrize("backend",... | I would rather update `test_raytracer` to add `pos=(0.0, 0.0, 2.0),` to `camera_1` instead of adding a new test? Because every single test we add is very costly (it requires building an entire scene), so it can be avoided, it is much better. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
pyrender_viewer = scene.visualizer.viewer._pyrender_viewer
assert pyrender_viewer.is_active
viewer_rgb, *_ = pyrender_viewer.render_offscreen(
pyrender_viewer._camera_node, pyrender_viewer._renderer, rgb=True, depth=False, seg=False, normal=False
)
# Render from add_camera
add_cam_rgb, ... | Why are you doing this? You need to have very strong motivations to add some xfail. Here you are basically saying that your fix does not work on windows. |
@pytest.mark.parametrize("renderer_type", [RENDERER_TYPE.RASTERIZER, RENDERER_TYPE.RAYTRACER])
def test_deformable_uv_textures(renderer_type, renderer, backend, show_viewer, png_snapshot):
CAM_RES = (256, 256)
# Increase tolerance for deformable rendering
png_snapshot.extension._std_err_threshold = 15.0
... | @@ -1543,3 +1543,31 @@ def test_add_camera_vs_interactive_viewer_consistency(add_box, renderer_type, sh
f"interactive viewer brightness ({viewer_brightness:.2f}), "
f"but ratio is {brightness_ratio:.2f}"
)
+
+
+@pytest.mark.required
+@pytest.mark.parametrize("renderer_type", [RENDERER_TYPE.RASTER... | tests/test_render.py | 26 | Python | 0.357 | bug | 153 | 49 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: tests/test_render.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: pyrender_viewer = scene.visualizer.viewer._pyrender_viewer
assert pyrender_viewer.is_active
viewer... | Why are you doing this? You need to have very strong motivations to add some xfail. Here you are basically saying that your fix does not work on windows. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# Compare brightness (mean pixel value)
viewer_brightness = viewer_rgb.mean()
add_cam_brightness = add_cam_rgb.mean()
brightness_ratio = add_cam_brightness / viewer_brightness
assert 0.99 <= brightness_ratio <= 1.01, (
f"add_camera brightness ({add_cam_brightness:.2f}) should match "
... | `viewer_options=gs.options.ViewerOptions(run_in_thread=False),` | dt=0.04,
substeps=6,
),
pbd_options=gs.options.PBDOptions(
particle_size=1e-2,
),
fem_options=gs.options.FEMOptions(
use_implicit_solver=True, # Implicit solver allows for larger timestep without failure on GPU backend
n_pcg_it... | @@ -1543,3 +1543,31 @@ def test_add_camera_vs_interactive_viewer_consistency(add_box, renderer_type, sh
f"interactive viewer brightness ({viewer_brightness:.2f}), "
f"but ratio is {brightness_ratio:.2f}"
)
+
+
+@pytest.mark.required
+@pytest.mark.parametrize("renderer_type", [RENDERER_TYPE.RASTER... | tests/test_render.py | 26 | Python | 0.5 | suggestion | 63 | 39 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: tests/test_render.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: # Compare brightness (mean pixel value)
viewer_brightness = viewer_rgb.mean()
add_cam_brightness =... | `viewer_options=gs.options.ViewerOptions(run_in_thread=False),` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
scene.build()
cam1.read()
cam2.read()
shared_metadata = cam1._shared_metadata
assert cam1._shared_metadata is cam2._shared_metadata
assert len(shared_metadata.sensors) == 2
assert shared_metadata.renderer is not None
scene.destroy()
assert shared_metadata.sensors is None
asse... | Could you check consistency with other camera explicitly? That would be much better. |
scene.build()
cam1.read()
cam2.read()
shared_metadata = cam1._shared_metadata
assert cam1._shared_metadata is cam2._shared_metadata
assert len(shared_metadata.sensors) == 2
assert shared_metadata.renderer is not None
scene.destroy()
assert shared_metadata.sensors is None
asse... | @@ -418,6 +418,31 @@ def test_raytracer_destroy():
assert shared_metadata.renderer is None
+@pytest.mark.required
+@pytest.mark.parametrize("backend", [gs.cuda])
+@pytest.mark.skipif(not ENABLE_RAYTRACER, reason="RayTracer is not supported because 'LuisaRenderPy' is not available.")
+def test_raytracer_attache... | tests/test_sensor_camera.py | 26 | Python | 0.286 | question | 84 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: tests/test_sensor_camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE:
scene.build()
cam1.read()
cam2.read()
shared_metadata = cam1._shared_metadata
ass... | Could you check consistency with other camera explicitly? That would be much better. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
import enum
import itertools
import os
import re
import sys
import time
import numpy as np
import pyglet
import pytest
import torch
import OpenGL.error
import genesis as gs
import genesis.utils.geom as gu
from genesis.utils import set_random_seed
from genesis.utils.image_exporter import FrameImageExporter, as_graysca... | Add a comment to explain that adding at least one entity is necessary to make sure the rendered image is not entirely blank, otherwise it is not possible to make sure something was actually rendered. |
# Render and verify
rgb_arr, *_ = camera.render(rgb=True)
rgb_arr = tensor_to_array(rgb_arr)
# Snapshot test for visual regression
assert rgb_array_to_png_bytes(rgb_arr) == png_snapshot
@pytest.mark.required
@pytest.mark.parametrize("renderer_type", [RENDERER_TYPE.RASTERIZER])
@pytest.mark.skipi... | @@ -1543,3 +1543,31 @@ def test_add_camera_vs_interactive_viewer_consistency(add_box, renderer_type, sh
f"interactive viewer brightness ({viewer_brightness:.2f}), "
f"but ratio is {brightness_ratio:.2f}"
)
+
+
+@pytest.mark.required
+@pytest.mark.parametrize("renderer_type", [RENDERER_TYPE.RASTER... | tests/test_render.py | 26 | Python | 0.357 | suggestion | 199 | 1,574 | 35 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: tests/test_render.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: import enum
import itertools
import os
import re
import sys
import time
import numpy as np
import pyglet
impo... | Add a comment to explain that adding at least one entity is necessary to make sure the rendered image is not entirely blank, otherwise it is not possible to make sure something was actually rendered. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
elif self._link is not None:
# Link exists but not built yet - use configured pose as-is (treat as world coordinates for now)
# This will be corrected when move_to_attach is called
pass
transform = pos_lookat_up_to_T(pos, lookat, up)
camera_wrapper = self._ge... | The `if` condition is useless since `self._register_camera_with_renderer()` is already an no-op if already registered. | # This will be corrected when move_to_attach is called
pass
transform = pos_lookat_up_to_T(pos, lookat, up)
self._camera_wrapper.transform = tensor_to_array(transform)
self._shared_metadata.renderer.update_camera(self._camera_wrapper)
def _apply_camera_transform(sel... | @@ -540,6 +559,10 @@ def _apply_camera_transform(self, camera_T: torch.Tensor):
def _render_current_state(self):
"""Perform the actual render for the current state."""
+ # Deferred registration for when using visualizer's rasterizer (visualizer now built)
+ if not self._camera_registered: | genesis/engine/sensors/camera.py | 26 | Python | 0.571 | suggestion | 118 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: elif self._link is not None:
# Link exists but not built yet - use configured ... | The `if` condition is useless since `self._register_camera_with_renderer()` is already an no-op if already registered. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._shared_metadata.renderer = Rasterizer(viewer=None, context=self._shared_metadata.context)
self._shared_metadata.renderer.build()
self._shared_metadata.sensors.append(self)
# Register camera now if standalone (offscreen), or defer to first render if using visualize... | While you are here, what is the point of `_get_camera_wrapper` ? This function seems useless. It was already the case because, but it is even worse now. Could you remove `_get_camera_wrapper` entirely and replace it direct access to `self._camera_wrapper` ? just add `self._register_camera_with_renderer()` on top of all... | self._shared_metadata.sensors.append(self)
# Register camera now if standalone (offscreen), or defer to first render if using visualizer's rasterizer
# (visualizer isn't built yet at sensor.build() time)
if self._shared_metadata.renderer.offscreen:
self._ensure_camera_regist... | @@ -431,29 +432,47 @@ def build(self):
self._shared_metadata.lights = gs.List()
self._shared_metadata.image_cache = {}
- # Create standalone rasterizer
- self._shared_metadata.context = self._create_standalone_context(scene)
-
- self._shared_metadata.renderer... | genesis/engine/sensors/camera.py | 26 | Python | 0.643 | suggestion | 360 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: self._shared_metadata.renderer = Rasterizer(viewer=None, context=self._shared_meta... | While you are here, what is the point of `_get_camera_wrapper` ? This function seems useless. It was already the case because, but it is even worse now. Could you remove `_get_camera_wrapper` entirely and replace it direct access to `self._camera_wrapper` ? just add `self._register_camera_with_renderer()` on top of all... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._camera_obj = visualizer.add_camera(
res=self._options.res,
pos=pos,
lookat=lookat,
up=up,
model=self._options.model,
fov=self._options.fov,
aperture=self._options.aperture,
focus_dist=self._options.focus_dist,
... | Just use `gu.trans_to_T`. It was introduced for this very use case. | spp=self._options.spp,
denoise=self._options.denoise,
near=0.05,
far=100.0,
env_idx=None if n_envs == 0 else 0,
debug=False,
)
# Attach the visualizer camera to the link if this sensor is attached
if self._link is not None:... | @@ -652,7 +675,12 @@ def build(self):
# Attach the visualizer camera to the link if this sensor is attached
if self._link is not None:
- offset_T = torch.tensor(self._options.offset_T, dtype=gs.tc_float, device=gs.device)
+ if self._options.offset_T is not None:
+ ... | genesis/engine/sensors/camera.py | 26 | Python | 0.429 | suggestion | 67 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: self._camera_obj = visualizer.add_camera(
res=self._options.res,
p... | Just use `gu.trans_to_T`. It was introduced for this very use case. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# This avoids OpenGL context conflicts since visualizer's rasterizer routes through viewer.render_offscreen()
visualizer = getattr(scene, "visualizer", None)
viewer = getattr(visualizer, "_viewer", None) if visualizer else None
if viewer is not None:
# Us... | Rename this function `_ensure_camera_registered` or anything that may indicate that it is safe to call this function repeatedly. Moreover, what is the point of adding `_with_renderer`? If you think it is useful you can keep it but I'm not sure it is adding much value here. | # If a viewer is active, reuse its windowed OpenGL context for both offscreen and onscreen
# rendering, rather than creating a separate headless context which is fragile.
if scene.viewer is not None:
self._shared_metadata.context = scene.visualizer.context
... | @@ -431,29 +432,47 @@ def build(self):
self._shared_metadata.lights = gs.List()
self._shared_metadata.image_cache = {}
- # Create standalone rasterizer
- self._shared_metadata.context = self._create_standalone_context(scene)
-
- self._shared_metadata.renderer... | genesis/engine/sensors/camera.py | 26 | Python | 0.643 | suggestion | 274 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: # This avoids OpenGL context conflicts since visualizer's rasterizer routes through vi... | Rename this function `_ensure_camera_registered` or anything that may indicate that it is safe to call this function repeatedly. Moreover, what is the point of adding `_with_renderer`? If you think it is useful you can keep it but I'm not sure it is adding much value here. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# ========================== Rasterizer Camera Sensor ==========================
@register_sensor(RasterizerCameraOptions, RasterizerCameraSharedMetadata, CameraData)
class RasterizerCameraSensor(BaseCameraSensor):
"""
Rasterizer camera sensor using OpenGL-based rendering.
This sensor renders RGB images... | Please use `is_` (or `has_`) prefix for boolean flags. |
# ========================== Rasterizer Camera Sensor ==========================
@register_sensor(RasterizerCameraOptions, RasterizerCameraSharedMetadata, CameraData)
class RasterizerCameraSensor(BaseCameraSensor):
"""
Rasterizer camera sensor using OpenGL-based rendering.
This sensor renders RGB image... | @@ -417,6 +417,7 @@ def __init__(
self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
+ self._camera_registered = False | genesis/engine/sensors/camera.py | 26 | Python | 0.429 | suggestion | 54 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE:
# ========================== Rasterizer Camera Sensor ==========================
@register_sens... | Please use `is_` (or `has_`) prefix for boolean flags. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
data_cls: Type[CameraData],
manager: "gs.SensorManager",
):
super().__init__(options, idx, data_cls, manager)
self._options: RasterizerCameraOptions
self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
self._camera_registered... | `scene.visualizer` is guaranteed to exist. No point to use `getattr` here. | idx: int,
data_cls: Type[CameraData],
manager: "gs.SensorManager",
):
super().__init__(options, idx, data_cls, manager)
self._options: RasterizerCameraOptions
self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
self.... | @@ -431,29 +432,47 @@ def build(self):
self._shared_metadata.lights = gs.List()
self._shared_metadata.image_cache = {}
- # Create standalone rasterizer
- self._shared_metadata.context = self._create_standalone_context(scene)
-
- self._shared_metadata.renderer... | genesis/engine/sensors/camera.py | 26 | Python | 0.429 | suggestion | 74 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: data_cls: Type[CameraData],
manager: "gs.SensorManager",
):
super().__... | `scene.visualizer` is guaranteed to exist. No point to use `getattr` here. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
manager: "gs.SensorManager",
):
super().__init__(options, idx, data_cls, manager)
self._options: RasterizerCameraOptions
self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
self._camera_registered = False
# ====================... | `visualizer._viewer` is guaranteed to exist. No point to use `getattr` here. Moreover, no need to access it via visualizer, you could get it directly from the scene, which is the expected public API:
```python
viewer = scene.viewer
``` | data_cls: Type[CameraData],
manager: "gs.SensorManager",
):
super().__init__(options, idx, data_cls, manager)
self._options: RasterizerCameraOptions
self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
self._is_camera_registe... | @@ -431,29 +432,47 @@ def build(self):
self._shared_metadata.lights = gs.List()
self._shared_metadata.image_cache = {}
- # Create standalone rasterizer
- self._shared_metadata.context = self._create_standalone_context(scene)
-
- self._shared_metadata.renderer... | genesis/engine/sensors/camera.py | 26 | Python | 0.857 | suggestion | 235 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: manager: "gs.SensorManager",
):
super().__init__(options, idx, data_cls, manag... | `visualizer._viewer` is guaranteed to exist. No point to use `getattr` here. Moreover, no need to access it via visualizer, you could get it directly from the scene, which is the expected public API:
```python
viewer = scene.viewer
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._camera_target = None
self._camera_wrapper = None
self._camera_registered = False
# ========================== Sensor Lifecycle ==========================
def build(self):
"""Initialize the rasterizer and register this camera."""
super().build()
scene = se... | I don't think it is relevant to define proxies for `scene.viewer` and `scene.visualizer`, just use them directly wherever to need to access them. | self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
self._is_camera_registered = False
# ========================== Sensor Lifecycle ==========================
def build(self):
"""Initialize the rasterizer and register this camera."""
... | @@ -431,29 +432,47 @@ def build(self):
self._shared_metadata.lights = gs.List()
self._shared_metadata.image_cache = {}
- # Create standalone rasterizer
- self._shared_metadata.context = self._create_standalone_context(scene)
-
- self._shared_metadata.renderer... | genesis/engine/sensors/camera.py | 26 | Python | 0.571 | suggestion | 145 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: self._camera_target = None
self._camera_wrapper = None
self._camera_regist... | I don't think it is relevant to define proxies for `scene.viewer` and `scene.visualizer`, just use them directly wherever to need to access them. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
self._options: RasterizerCameraOptions
self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
self._camera_registered = False
# ========================== Sensor Lifecycle ==========================
def build(self):
"""Initialize the ras... | The point is more about the fact that the same windowed OpenGL context will be reused, which is capable of handling both offscreen and onscreen rendering, whereas managing two contexts, with one being headless and the other windowed, is proven to be very painful and fragile. | super().__init__(options, idx, data_cls, manager)
self._options: RasterizerCameraOptions
self._camera_node = None
self._camera_target = None
self._camera_wrapper = None
self._is_camera_registered = False
# ========================== Sensor Lifecycle =================... | @@ -431,29 +432,47 @@ def build(self):
self._shared_metadata.lights = gs.List()
self._shared_metadata.image_cache = {}
- # Create standalone rasterizer
- self._shared_metadata.context = self._create_standalone_context(scene)
-
- self._shared_metadata.renderer... | genesis/engine/sensors/camera.py | 26 | Python | 0.5 | suggestion | 275 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE: self._options: RasterizerCameraOptions
self._camera_node = None
self._came... | The point is more about the fact that the same windowed OpenGL context will be reused, which is capable of handling both offscreen and onscreen rendering, whereas managing two contexts, with one being headless and the other windowed, is proven to be very painful and fragile. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
transform = pos_lookat_up_to_T(pos, lookat, up)
self._camera_wrapper.transform = tensor_to_array(transform)
self._shared_metadata.renderer.update_camera(self._camera_wrapper)
def _apply_camera_transform(self, camera_T: torch.Tensor):
"""Update rasterizer camera wrapper from a world... | The CI is currently failing with the error:
```python
> rgb_tensor = torch.from_numpy(rgb_arr).to(dtype=torch.uint8, device=gs.device)
E ValueError: At least one stride in the given numpy array is negative, and tensors with negative strides are not currently supported. (You can probably work around this by ... |
transform = pos_lookat_up_to_T(pos, lookat, up)
self._camera_wrapper.transform = tensor_to_array(transform)
self._shared_metadata.renderer.update_camera(self._camera_wrapper)
def _apply_camera_transform(self, camera_T: torch.Tensor):
"""Update rasterizer camera wrapper from a world... | @@ -522,35 +539,30 @@ def _update_camera_pose(self):
pass
transform = pos_lookat_up_to_T(pos, lookat, up)
- camera_wrapper = self._get_camera_wrapper()
- camera_wrapper.transform = tensor_to_array(transform)
- self._shared_metadata.renderer.update_camera(camera_wrapper)
-
- ... | genesis/engine/sensors/camera.py | 26 | Python | 1 | bug | 551 | 51 | 51 | false | [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer. | 2,351 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/sensors/camera.py
PR Title: [BUG FIX] Fix OpenGL context clash when combining Rasterizer-based camera sensors and interactive viewer.
BEFORE CODE:
transform = pos_lookat_up_to_T(pos, lookat, up)
self._camera_wrapper.transform = ... | The CI is currently failing with the error:
```python
> rgb_tensor = torch.from_numpy(rgb_arr).to(dtype=torch.uint8, device=gs.device)
E ValueError: At least one stride in the given numpy array is negative, and tensors with negative strides are not currently supported. (You can probably work around this by ... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
gs.JOINT_TYPE.REVOLUTE: (("angular",),),
gs.JOINT_TYPE.PRISMATIC: (("linear",),),
gs.JOINT_TYPE.SPHERICAL: (
("rotX",),
("rotY",),
("rotZ",),
),
gs.JOINT_TYPE.FIXED: (),
gs.JOINT_TYPE.FREE: (
("transX",),
("transY",),
("transZ",),
("rotX",)... | Just do:
```python
joint_pos = usd_pos_to_numpy(joint_pos_attr.Get()) if joint_pos_attr.HasValue() else gu.zero_pos()
``` | DRIVE_NAMES = {
gs.JOINT_TYPE.REVOLUTE: (("angular",),),
gs.JOINT_TYPE.PRISMATIC: (("linear",),),
gs.JOINT_TYPE.SPHERICAL: (
("rotX",),
("rotY",),
("rotZ",),
),
gs.JOINT_TYPE.FIXED: (),
gs.JOINT_TYPE.FREE: (
("transX",),
("transY",),
("transZ",),
... | @@ -40,7 +44,10 @@
def _parse_joint_axis_pos(
context: UsdContext, joint: UsdPhysics.Joint, child_link: Usd.Prim, is_body1: bool
) -> Tuple[str, np.ndarray, np.ndarray]:
- joint_pos = usd_pos_to_numpy((joint.GetLocalPos1Attr() if is_body1 else joint.GetLocalPos0Attr()).Get())
+ joint_pos_attr = joint.GetLo... | genesis/utils/usd/usd_rigid_entity.py | 26 | Python | 0.714 | suggestion | 121 | 51 | 51 | false | [BUG FIX] Fix handling of unspecified spatial inertia properties in USD. | 2,401 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alanray-tech | Review this pull request change.
Language: Python
File: genesis/utils/usd/usd_rigid_entity.py
PR Title: [BUG FIX] Fix handling of unspecified spatial inertia properties in USD.
BEFORE CODE: gs.JOINT_TYPE.REVOLUTE: (("angular",),),
gs.JOINT_TYPE.PRISMATIC: (("linear",),),
gs.JOINT_TYPE.SPHERICAL: (
... | Just do:
```python
joint_pos = usd_pos_to_numpy(joint_pos_attr.Get()) if joint_pos_attr.HasValue() else gu.zero_pos()
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
l_info["invweight"] = np.full((2,), fill_value=-1.0)
# parse link fixed state
link_fixed = False
if link.HasAPI(UsdPhysics.RigidBodyAPI):
rigid_body_api = UsdPhysics.RigidBodyAPI(link)
if rigid_body_api.GetKinematicEnabledAttr().Get():
link_fixed = True
elif rigid_bo... | Just implement `usd_inertia_to_numpy(inertia_attr.Get())` that calls `np.diag` internally. |
# parse link fixed state
link_fixed = False
if link.HasAPI(UsdPhysics.RigidBodyAPI):
rigid_body_api = UsdPhysics.RigidBodyAPI(link)
if rigid_body_api.GetKinematicEnabledAttr().Get():
link_fixed = True
elif rigid_body_api.GetRigidBodyEnabledAttr().Get() is False:
... | @@ -90,10 +97,21 @@ def _parse_link(
# parse link mass properties
if link.HasAPI(UsdPhysics.MassAPI):
mass_api = UsdPhysics.MassAPI(link)
- l_info["inertial_pos"] = usd_pos_to_numpy(mass_api.GetCenterOfMassAttr().Get())
- l_info["inertial_quat"] = usd_quat_to_numpy(mass_api.GetPrincipal... | genesis/utils/usd/usd_rigid_entity.py | 26 | Python | 0.357 | suggestion | 90 | 51 | 51 | false | [BUG FIX] Fix handling of unspecified spatial inertia properties in USD. | 2,401 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alanray-tech | Review this pull request change.
Language: Python
File: genesis/utils/usd/usd_rigid_entity.py
PR Title: [BUG FIX] Fix handling of unspecified spatial inertia properties in USD.
BEFORE CODE: l_info["invweight"] = np.full((2,), fill_value=-1.0)
# parse link fixed state
link_fixed = False
if link.HasA... | Just implement `usd_inertia_to_numpy(inertia_attr.Get())` that calls `np.diag` internally. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
if rigid_body_api.GetKinematicEnabledAttr().Get():
link_fixed = True
elif rigid_body_api.GetRigidBodyEnabledAttr().Get() is False:
link_fixed = True
elif link.HasAPI(UsdPhysics.CollisionAPI):
link_fixed = True
if morph.fixed:
link_fixed = any(parent_idx =... | `l_info["inertial_i"] = usd_inertia_to_numpy(inertia_attr.Get())` | link_fixed = True
elif rigid_body_api.GetRigidBodyEnabledAttr().Get() is False:
link_fixed = True
elif link.HasAPI(UsdPhysics.CollisionAPI):
link_fixed = True
if morph.fixed:
link_fixed = any(parent_idx == -1 for _, parent_idx, _ in joints)
# parse link mass... | @@ -90,10 +97,21 @@ def _parse_link(
# parse link mass properties
if link.HasAPI(UsdPhysics.MassAPI):
mass_api = UsdPhysics.MassAPI(link)
- l_info["inertial_pos"] = usd_pos_to_numpy(mass_api.GetCenterOfMassAttr().Get())
- l_info["inertial_quat"] = usd_quat_to_numpy(mass_api.GetPrincipal... | genesis/utils/usd/usd_rigid_entity.py | 26 | Python | 0.357 | suggestion | 65 | 51 | 51 | false | [BUG FIX] Fix handling of unspecified spatial inertia properties in USD. | 2,401 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alanray-tech | Review this pull request change.
Language: Python
File: genesis/utils/usd/usd_rigid_entity.py
PR Title: [BUG FIX] Fix handling of unspecified spatial inertia properties in USD.
BEFORE CODE: if rigid_body_api.GetKinematicEnabledAttr().Get():
link_fixed = True
elif rigid_body_api.GetRigidB... | `l_info["inertial_i"] = usd_inertia_to_numpy(inertia_attr.Get())` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
def usd_diagonal_inertia_to_numpy(usd_pos: Gf.Vec3f) -> np.ndarray | None:
"""
Convert USD diagonal inertia to numpy array, handling default and invalid values.
The USD Physics MassAPI defines diagonalInertia with default value (0, 0, 0),
which is valid but means the inertia should be ignored/computed... | I would rather check `all(math.isfinite(e) for e in inertia)`. Because we do not want to hind random negative values. |
def usd_inertia_to_numpy(inertia: Gf.Vec3f) -> np.ndarray | None:
"""
Convert USD diagonal inertia to numpy diagonal matrix, handling default and invalid values.
The USD Physics MassAPI defines diagonalInertia with default value (0, 0, 0),
which is valid but means the inertia should be ignored/compute... | @@ -25,18 +26,151 @@
}
-def usd_pos_to_numpy(usd_pos: Gf.Vec3f) -> np.ndarray:
+def usd_pos_to_numpy(usd_pos: Gf.Vec3f | None) -> np.ndarray:
+ """
+ Convert USD position to numpy array, handling None values.
+
+ Parameters
+ ----------
+ usd_pos : Gf.Vec3f | None
+ USD position attribute val... | genesis/utils/usd/usd_utils.py | 26 | Python | 0.571 | suggestion | 117 | 51 | 51 | false | [BUG FIX] Fix handling of unspecified spatial inertia properties in USD. | 2,401 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alanray-tech | Review this pull request change.
Language: Python
File: genesis/utils/usd/usd_utils.py
PR Title: [BUG FIX] Fix handling of unspecified spatial inertia properties in USD.
BEFORE CODE:
def usd_diagonal_inertia_to_numpy(usd_pos: Gf.Vec3f) -> np.ndarray | None:
"""
Convert USD diagonal inertia to numpy array, ... | I would rather check `all(math.isfinite(e) for e in inertia)`. Because we do not want to hind random negative values. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
if usd_mass <= 0:
return None
# Check for invalid values (inf or nan)
if np.isinf(usd_mass) or np.isnan(usd_mass):
return None
return float(usd_mass)
def usd_attr_array_to_numpy(attr: Usd.Attribute, dtype: np.dtype, return_none: bool = False) -> np.ndarray | None:
if attr.HasValue(... | Do not do this. It hides user-specified invalid values. | if not all(math.isfinite(e) for e in inertia):
return None
return inertia
def usd_mass_to_float(usd_mass: float) -> float | None:
"""
Convert USD mass to float, handling default and invalid values.
The USD Physics MassAPI defines mass with default value 0, which is valid but
means the... | @@ -25,18 +26,151 @@
}
-def usd_pos_to_numpy(usd_pos: Gf.Vec3f) -> np.ndarray:
+def usd_pos_to_numpy(usd_pos: Gf.Vec3f | None) -> np.ndarray:
+ """
+ Convert USD position to numpy array, handling None values.
+
+ Parameters
+ ----------
+ usd_pos : Gf.Vec3f | None
+ USD position attribute val... | genesis/utils/usd/usd_utils.py | 26 | Python | 0.286 | suggestion | 55 | 44 | 51 | false | [BUG FIX] Fix handling of unspecified spatial inertia properties in USD. | 2,401 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | alanray-tech | Review this pull request change.
Language: Python
File: genesis/utils/usd/usd_utils.py
PR Title: [BUG FIX] Fix handling of unspecified spatial inertia properties in USD.
BEFORE CODE: if usd_mass <= 0:
return None
# Check for invalid values (inf or nan)
if np.isinf(usd_mass) or np.isnan(usd_mass)... | Do not do this. It hides user-specified invalid values. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
import os
from frozendict import frozendict
import quadrants as qd
import numpy as np
import torch
import genesis as gs
import genesis.utils.geom as gu
import genesis.utils.array_class as array_class
from genesis.engine.solvers.rigid.abd import func_solve_mass_batch
from genesis.utils.misc import qd_to_torch
from . ... | Why is this optional? I'm not a fan of this kind of hidden magic flag. Only add them if necessary. | from typing import TYPE_CHECKING
import numpy as np
import quadrants as qd
import torch
from frozendict import frozendict
import genesis as gs
import genesis.utils.array_class as array_class
import genesis.utils.geom as gu
from genesis.engine.solvers.rigid.abd import func_solve_mass_batch
from genesis.utils.misc impo... | @@ -20,6 +23,8 @@
IS_OLD_TORCH = tuple(map(int, torch.__version__.split(".")[:2])) < (2, 8)
+GS_SOLVER_DECOMPOSE = os.environ.get("GS_SOLVER_DECOMPOSE", "1") == "1"
+ | genesis/engine/solvers/rigid/constraint/solver.py | 26 | Python | 0.214 | question | 98 | 51 | 51 | false | [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations. | 2,406 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: genesis/engine/solvers/rigid/constraint/solver.py
PR Title: [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations.
BEFORE CODE: import os
from frozendict import frozendict
import quadrants ... | Why is this optional? I'm not a fan of this kind of hidden magic flag. Only add them if necessary. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
constraint_state=constraint_state,
static_rigid_sim_config=static_rigid_sim_config,
)
@ti.kernel(fastcache=gs.use_fastcache)
def _kernel_update_search_direction(
constraint_state: array_class.ConstraintState,
rigid_global_info: array_class.RigidGlobalInfo,
stati... | Same. Remove this registration function if it can be avoid. | constraint_state=constraint_state,
static_rigid_sim_config=static_rigid_sim_config,
)
@ti.kernel(fastcache=gs.use_fastcache)
def _kernel_update_search_direction(
constraint_state: array_class.ConstraintState,
rigid_global_info: array_class.RigidGlobalInfo,
stati... | @@ -0,0 +1,191 @@
+import quadrants as ti
+
+import genesis as gs
+import genesis.utils.array_class as array_class
+from genesis.engine.solvers.rigid.constraint import solver
+
+
+@ti.kernel(fastcache=gs.use_fastcache)
+def _kernel_linesearch(
+ entities_info: array_class.EntitiesInfo,
+ dofs_state: array_class.D... | genesis/engine/solvers/rigid/constraint/solver_breakdown.py | 26 | Python | 0.286 | refactor | 59 | 51 | 51 | false | [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations. | 2,406 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: genesis/engine/solvers/rigid/constraint/solver_breakdown.py
PR Title: [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations.
BEFORE CODE: constraint_state=constraint_state,
... | Same. Remove this registration function if it can be avoid. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@ti.kernel(fastcache=gs.use_fastcache)
def _kernel_update_search_direction(
constraint_state: array_class.ConstraintState,
rigid_global_info: array_class.RigidGlobalInfo,
static_rigid_sim_config: ti.template(),
):
"""Step 6: Check convergence and update search direction"""
_B = constraint_state.gr... | Why '_macrokernels' suffix? This function is local to this file, just call it `func_solve_decomposed`. |
@ti.kernel(fastcache=gs.use_fastcache)
def _kernel_update_search_direction(
constraint_state: array_class.ConstraintState,
rigid_global_info: array_class.RigidGlobalInfo,
static_rigid_sim_config: ti.template(),
):
"""Step 6: Check convergence and update search direction"""
_B = constraint_state.gr... | @@ -0,0 +1,191 @@
+import quadrants as ti
+
+import genesis as gs
+import genesis.utils.array_class as array_class
+from genesis.engine.solvers.rigid.constraint import solver
+
+
+@ti.kernel(fastcache=gs.use_fastcache)
+def _kernel_linesearch(
+ entities_info: array_class.EntitiesInfo,
+ dofs_state: array_class.D... | genesis/engine/solvers/rigid/constraint/solver_breakdown.py | 26 | Python | 0.5 | refactor | 102 | 51 | 51 | false | [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations. | 2,406 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: genesis/engine/solvers/rigid/constraint/solver_breakdown.py
PR Title: [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations.
BEFORE CODE:
@ti.kernel(fastcache=gs.use_fastcache)
def _kernel... | Why '_macrokernels' suffix? This function is local to this file, just call it `func_solve_decomposed`. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
constraint_state=constraint_state,
static_rigid_sim_config=static_rigid_sim_config,
)
@ti.kernel(fastcache=gs.use_fastcache)
def _kernel_update_search_direction(
constraint_state: array_class.ConstraintState,
rigid_global_info: array_class.RigidGlobalInfo,
stati... | Remove print. Using logger at debug level. | constraint_state=constraint_state,
static_rigid_sim_config=static_rigid_sim_config,
)
@ti.kernel(fastcache=gs.use_fastcache)
def _kernel_update_search_direction(
constraint_state: array_class.ConstraintState,
rigid_global_info: array_class.RigidGlobalInfo,
stati... | @@ -0,0 +1,191 @@
+import quadrants as ti
+
+import genesis as gs
+import genesis.utils.array_class as array_class
+from genesis.engine.solvers.rigid.constraint import solver
+
+
+@ti.kernel(fastcache=gs.use_fastcache)
+def _kernel_linesearch(
+ entities_info: array_class.EntitiesInfo,
+ dofs_state: array_class.D... | genesis/engine/solvers/rigid/constraint/solver_breakdown.py | 26 | Python | 0.286 | bug | 42 | 51 | 51 | false | [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations. | 2,406 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: genesis/engine/solvers/rigid/constraint/solver_breakdown.py
PR Title: [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations.
BEFORE CODE: constraint_state=constraint_state,
... | Remove print. Using logger at debug level. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
entities_info=entities_info,
rigid_global_info=rigid_global_info,
constraint_state=constraint_state,
static_rigid_sim_config=static_rigid_sim_config,
)
func_terminate_or_update_descent_batch(
i_b,
constraint_state=constraint_state,... | `_monolith` is not clear to me. What are you referring to with this keyword? | static_rigid_sim_config=static_rigid_sim_config,
)
func_terminate_or_update_descent_batch(
i_b,
constraint_state=constraint_state,
rigid_global_info=rigid_global_info,
static_rigid_sim_config=static_rigid_sim_config,
)
@qd.perf_dispa... | @@ -2938,13 +2998,30 @@ def func_solve_iter(
)
-@qd.kernel(fastcache=gs.use_fastcache)
+@qd.perf_dispatch(
+ get_geometry_hash=lambda *args, **kwargs: (*args, frozendict(kwargs)), warmup=3, active=3, repeat_after_seconds=1.0
+)
def func_solve_body(
entities_info: array_class.EntitiesInfo,
dof... | genesis/engine/solvers/rigid/constraint/solver.py | 26 | Python | 0.357 | question | 76 | 51 | 51 | false | [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations. | 2,406 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: genesis/engine/solvers/rigid/constraint/solver.py
PR Title: [MISC] Decompose solver and use parallel Hessian and Cholesky to significantly speedup contact-rich and dynamic simulations.
BEFORE CODE: entities_info=entities_info,
rigid_glo... | `_monolith` is not clear to me. What are you referring to with this keyword? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
Note that, for completeness, Genesis code does handle resetting errno to 0, inside set_qpos, but for simplicity,
we make resetting errno explicit in this test.
"""
import copy
import importlib.util
import xml.etree.ElementTree as ET
from pathlib import Path
from typing import TYPE_CHECKING, Callable, cast
import num... | Could you remove "GJK is iterative so needs more slack.". I don't think it is true. |
Note that, for completeness, Genesis code does handle resetting errno to 0, inside set_qpos, but for simplicity,
we make resetting errno explicit in this test.
"""
import copy
import importlib.util
import xml.etree.ElementTree as ET
from pathlib import Path
from typing import TYPE_CHECKING, Callable, cast
import num... | @@ -49,6 +49,49 @@
ERRNO_CALLED_GJK = 1 << 16
POS_TOL = 1e-2 # otherwise tests fail
+# Tolerances for checking results against hand-computed expected values.
+# Analytical solutions should be near-exact; GJK is iterative so needs more slack. | tests/test_rigid_physics_analytical_vs_gjk.py | 26 | Python | 0.286 | suggestion | 83 | 51 | 51 | false | [MISC] Add expected values for gjk vs analytical collisions | 2,413 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_physics_analytical_vs_gjk.py
PR Title: [MISC] Add expected values for gjk vs analytical collisions
BEFORE CODE:
Note that, for completeness, Genesis code does handle resetting errno to 0, inside set_qpos, but for simplicity,
we make resetting... | Could you remove "GJK is iterative so needs more slack.". I don't think it is true. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
import xml.etree.ElementTree as ET
from pathlib import Path
from typing import TYPE_CHECKING, Callable, cast
import numpy as np
import pytest
import genesis as gs
from .utils import assert_allclose
if TYPE_CHECKING:
from genesis.engine.entities.rigid_entity import RigidGeom
ERRNO_CALLED_GJK = 1 << 16
POS_TOL =... | You could use `TOL_SINGLE` for this? It is designed for this.
`from .conftest import TOL_SINGLE` | import xml.etree.ElementTree as ET
from pathlib import Path
from typing import TYPE_CHECKING, Callable, cast
import numpy as np
import pytest
import genesis as gs
from .utils import assert_allclose
from .conftest import TOL_SINGLE
if TYPE_CHECKING:
from genesis.engine.entities.rigid_entity import RigidGeom
ERR... | @@ -49,6 +49,49 @@
ERRNO_CALLED_GJK = 1 << 16
POS_TOL = 1e-2 # otherwise tests fail
+# Tolerances for checking results against hand-computed expected values.
+# Analytical solutions should be near-exact; GJK is iterative so needs more slack.
+#
+# Penetration tolerance: absolute error in metres.
+# Normal toleranc... | tests/test_rigid_physics_analytical_vs_gjk.py | 26 | Python | 0.429 | suggestion | 97 | 51 | 51 | false | [MISC] Add expected values for gjk vs analytical collisions | 2,413 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_physics_analytical_vs_gjk.py
PR Title: [MISC] Add expected values for gjk vs analytical collisions
BEFORE CODE: import xml.etree.ElementTree as ET
from pathlib import Path
from typing import TYPE_CHECKING, Callable, cast
import numpy as np
im... | You could use `TOL_SINGLE` for this? It is designed for this.
`from .conftest import TOL_SINGLE` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
os.environ["EGL_DEVICE_ID"] = str(_get_egl_index(gpu_index))
except (AttributeError, KeyError):
# AttributeError: CUDA is not supported on this machine
# KeyError: The selected GPU does not support CUDA
pass
def pytest_add... | Are you sure it is really necessary to expose this many options?
I have the feeling just adding `--profile-num-steps` that would be equal to 0 by default should be good enough?
| if has_egl:
try:
os.environ["EGL_DEVICE_ID"] = str(_get_egl_index(gpu_index))
except (AttributeError, KeyError):
# AttributeError: CUDA is not supported on this machine
# KeyError: The selected GPU does not support C... | @@ -416,13 +416,68 @@ def pytest_addoption(parser):
else SUPPRESS
)
parser.addoption("--mem-monitoring-filepath", type=str, help=help_text)
+ parser.addoption("--profile", action="store_true", default=False, help="Enable PyTorch profiling for benchmarks.")
+ parser.addoption("--profile-wait", t... | tests/conftest.py | 26 | Python | 0.571 | question | 177 | 51 | 51 | false | [MISC] Add torch profile facilities to rigid benchmarks. | 2,421 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/conftest.py
PR Title: [MISC] Add torch profile facilities to rigid benchmarks.
BEFORE CODE: os.environ["EGL_DEVICE_ID"] = str(_get_egl_index(gpu_index))
except (AttributeError, KeyError):
# Attribut... | Are you sure it is really necessary to expose this many options?
I have the feeling just adding `--profile-num-steps` that would be equal to 0 by default should be good enough?
| You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
parser.addoption("--vis", action="store_true", default=False, help="Enable interactive viewer.")
parser.addoption("--dev", action="store_true", default=False, help="Enable genesis debug mode.")
supported, _reason = is_mem_monitoring_supported()
help_text = (
"Run memory monitoring, and store... | Use nullcontext; See `examples/speed_benchmark/timers.py` | parser.addoption("--vis", action="store_true", default=False, help="Enable interactive viewer.")
parser.addoption("--dev", action="store_true", default=False, help="Enable genesis debug mode.")
supported, _reason = is_mem_monitoring_supported()
help_text = (
"Run memory monitoring, and store... | @@ -416,13 +418,25 @@ def pytest_addoption(parser):
else SUPPRESS
)
parser.addoption("--mem-monitoring-filepath", type=str, help=help_text)
+ if os.environ.get("GS_PROFILING", "0") == "1":
+ profiling.parser_add_options(parser)
@pytest.fixture(scope="session")
def show_viewer(pytestco... | tests/conftest.py | 26 | Python | 0.571 | suggestion | 57 | 51 | 51 | false | [MISC] Add torch profile facilities to rigid benchmarks. | 2,421 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/conftest.py
PR Title: [MISC] Add torch profile facilities to rigid benchmarks.
BEFORE CODE: parser.addoption("--vis", action="store_true", default=False, help="Enable interactive viewer.")
parser.addoption("--dev", action="store_true", defaul... | Use nullcontext; See `examples/speed_benchmark/timers.py` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
wait = pytestconfig.getoption("--profile-wait")
warmup = pytestconfig.getoption("--profile-warmup")
active = pytestconfig.getoption("--profile-active")
repeat = pytestconfig.getoption("--profile-repeat")
schedule = torch.profiler.schedule(wait=wait, warmup=warmup, active=active, repeat=repeat)
... | I don't get it. What not just use the context as it is designed to be used instead of calling __enter__ and __exit__ manually? | warmup = pytestconfig.getoption("--profile-warmup")
active = pytestconfig.getoption("--profile-active")
repeat = pytestconfig.getoption("--profile-repeat")
schedule = torch.profiler.schedule(wait=wait, warmup=warmup, active=active, repeat=repeat)
activities = [ProfilerActivity.CPU]
if torch.cu... | @@ -0,0 +1,71 @@
+import argparse
+import pytest
+
+
+def parser_add_options(parser: pytest.Parser) -> None:
+ parser.addoption(
+ "--profile-wait",
+ type=int,
+ required=True,
+ help="Number of steps to wait before profiling. Depends on what you want to view, since the profile will like... | tests/profiling.py | 26 | Python | 0.429 | question | 126 | 31 | 27 | false | [MISC] Add torch profile facilities to rigid benchmarks. | 2,421 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/profiling.py
PR Title: [MISC] Add torch profile facilities to rigid benchmarks.
BEFORE CODE: wait = pytestconfig.getoption("--profile-wait")
warmup = pytestconfig.getoption("--profile-warmup")
active = pytestconfig.getoption("--profile-active... | I don't get it. What not just use the context as it is designed to be used instead of calling __enter__ and __exit__ manually? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
def write(self, items):
nonlocal stream_writers
if self.wandb_run is not None:
self.wandb_run.log(
{
"timestamp": self.wandb_run.config["timestamp"],
**items,
}
)
... | Why not to yield only 'profiler_step' since the other argument is used?
I would recommend doing this and extending the API later if necessary. | def write(self, items):
nonlocal stream_writers
if self.wandb_run is not None:
self.wandb_run.log(
{
"timestamp": self.wandb_run.config["timestamp"],
**items,
}
)
... | @@ -313,7 +313,8 @@ def write(self, items):
@pytest.fixture
-def go2(solver, n_envs, gjk):
+def go2(solver, n_envs, gjk, pytorch_profiler):
+ _, profiler_step = pytorch_profiler | tests/test_rigid_benchmarks.py | 26 | Python | 0.429 | question | 143 | 51 | 51 | false | [MISC] Add torch profile facilities to rigid benchmarks. | 2,421 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add torch profile facilities to rigid benchmarks.
BEFORE CODE: def write(self, items):
nonlocal stream_writers
if self.wandb_run is not None:
self.wandb_run.log(
... | Why not to yield only 'profiler_step' since the other argument is used?
I would recommend doing this and extending the API later if necessary. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
parser.addoption("--vis", action="store_true", default=False, help="Enable interactive viewer.")
parser.addoption("--dev", action="store_true", default=False, help="Enable genesis debug mode.")
supported, _reason = is_mem_monitoring_supported()
help_text = (
"Run memory monitoring, and store... | Why not to move this fixture in profiling since you already created a dedicated file? It would not work? | parser.addoption("--vis", action="store_true", default=False, help="Enable interactive viewer.")
parser.addoption("--dev", action="store_true", default=False, help="Enable genesis debug mode.")
supported, _reason = is_mem_monitoring_supported()
help_text = (
"Run memory monitoring, and store... | @@ -416,13 +418,25 @@ def pytest_addoption(parser):
else SUPPRESS
)
parser.addoption("--mem-monitoring-filepath", type=str, help=help_text)
+ if os.environ.get("GS_PROFILING", "0") == "1":
+ profiling.parser_add_options(parser)
@pytest.fixture(scope="session")
def show_viewer(pytestco... | tests/conftest.py | 26 | Python | 0.429 | refactor | 104 | 51 | 51 | false | [MISC] Add torch profile facilities to rigid benchmarks. | 2,421 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/conftest.py
PR Title: [MISC] Add torch profile facilities to rigid benchmarks.
BEFORE CODE: parser.addoption("--vis", action="store_true", default=False, help="Enable interactive viewer.")
parser.addoption("--dev", action="store_true", defaul... | Why not to move this fixture in profiling since you already created a dedicated file? It would not work? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
def write(self, items):
nonlocal stream_writers
if self.wandb_run is not None:
self.wandb_run.log(
{
"timestamp": self.wandb_run.config["timestamp"],
**items,
}
)
... | I hate just renaming variables. Just call your fixture `pytorch_profiler_step`, or `profiler_calling` or something. | def write(self, items):
nonlocal stream_writers
if self.wandb_run is not None:
self.wandb_run.log(
{
"timestamp": self.wandb_run.config["timestamp"],
**items,
}
)
... | @@ -313,7 +313,8 @@ def write(self, items):
@pytest.fixture
-def go2(solver, n_envs, gjk):
+def go2(solver, n_envs, gjk, pytorch_profiler):
+ profiler_step = pytorch_profiler | tests/test_rigid_benchmarks.py | 26 | Python | 0.5 | style | 115 | 51 | 51 | false | [MISC] Add torch profile facilities to rigid benchmarks. | 2,421 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | hughperkins | Review this pull request change.
Language: Python
File: tests/test_rigid_benchmarks.py
PR Title: [MISC] Add torch profile facilities to rigid benchmarks.
BEFORE CODE: def write(self, items):
nonlocal stream_writers
if self.wandb_run is not None:
self.wandb_run.log(
... | I hate just renaming variables. Just call your fixture `pytorch_profiler_step`, or `profiler_calling` or something. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# Pre-compute the entity set for rigid state retrieval
self._rigid_retrieve_entity_set = set(
self._entities_by_coupling_type["two_way_soft_constraint"]
+ self._entities_by_coupling_type["ipc_only"]
+ self._articulation_with_non_fixed_base
)
# Pre-al... | Just do ` workspace = os.path.join(tempfile.gettempdir(), f"genesis_ipc_{self.sim.scene.uid.full()}")` | # Pre-compute the entity set for rigid state retrieval
self._rigid_retrieve_entity_set = set(
self._entities_by_coupling_type["two_way_soft_constraint"]
+ self._entities_by_coupling_type["ipc_only"]
+ self._articulation_with_non_fixed_base
)
# Pre-all... | @@ -316,7 +313,9 @@ def _init_ipc(self):
UIPCTimer.disable_all()
# Create workspace directory for IPC output, named after scene UID.
- workspace = os.path.join(tempfile.gettempdir(), f"genesis_ipc_{self.sim.scene.uid}")
+ # The UID contains < and > which are illegal in Windows path... | genesis/engine/couplers/ipc_coupler/coupler.py | 26 | Python | 0.5 | suggestion | 102 | 51 | 51 | false | [MISC] Remove partially broken IPC contact proxy option. | 2,463 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | Roushelfy | Review this pull request change.
Language: Python
File: genesis/engine/couplers/ipc_coupler/coupler.py
PR Title: [MISC] Remove partially broken IPC contact proxy option.
BEFORE CODE:
# Pre-compute the entity set for rigid state retrieval
self._rigid_retrieve_entity_set = set(
self._enti... | Just do ` workspace = os.path.join(tempfile.gettempdir(), f"genesis_ipc_{self.sim.scene.uid.full()}")` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
import genesis as gs
gs.init()
scene = gs.Scene(
show_viewer=True,
viewer_options=gs.options.ViewerOptions(
res=(1280, 960),
camera_pos=(3.5, 0.0, 2.5),
camera_lookat=(0.0, 0.0, 0.5),
camera_fov=50,
),
vis_options=gs.options.VisOptions(
rendered_envs_idx=list(ra... | No issues found. | import genesis as gs
gs.init()
scene = gs.Scene(
show_viewer=True,
viewer_options=gs.options.ViewerOptions(
res=(1280, 960),
camera_pos=(3.5, 0.0, 2.5),
camera_lookat=(0.0, 0.0, 0.5),
camera_fov=50,
),
vis_options=gs.options.VisOptions(
rendered_envs_idx=list(ra... | @@ -0,0 +1,30 @@
+import genesis as gs
+
+gs.init()
+
+scene = gs.Scene(
+ show_viewer=True,
+ viewer_options=gs.options.ViewerOptions(
+ res=(1280, 960),
+ camera_pos=(3.5, 0.0, 2.5),
+ camera_lookat=(0.0, 0.0, 0.5),
+ camera_fov=50,
+ ),
+ vis_options=gs.options.VisOptions(
+ ... | examples/tutorials/selecting_rendered_envs.py | 0 | Python | 1 | none | 16 | 31 | 31 | true | [FEATURE] Modify n_rendered_envs to rendered_envs_idx | 752 | Genesis-Embodied-AI/Genesis | 26,702 | Python | lupinjia | Review this pull request change.
Language: Python
File: examples/tutorials/selecting_rendered_envs.py
PR Title: [FEATURE] Modify n_rendered_envs to rendered_envs_idx
BEFORE CODE: import genesis as gs
gs.init()
scene = gs.Scene(
show_viewer=True,
viewer_options=gs.options.ViewerOptions(
res=(1280, ... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
for j_d in range(i_d + 1, self._solver.n_dofs):
dot = gs.ti_float(0.0)
for k_d in range(i_d):
dot = dot + self.nt_H[j_d, k_d, i_b] * self.nt_H[i_d, k_d, i_b]
self.nt_H[j_d, i_d, i_b] = (self.nt_H[j_d, i_d, i_b] - dot) * tmp
@ti.func
... | No issues found. | for j_d in range(i_d + 1, self._solver.n_dofs):
dot = gs.ti_float(0.0)
for k_d in range(i_d):
dot = dot + self.nt_H[j_d, k_d, i_b] * self.nt_H[i_d, k_d, i_b]
self.nt_H[j_d, i_d, i_b] = (self.nt_H[j_d, i_d, i_b] - dot) * tmp
@ti.func
... | @@ -191,103 +191,169 @@ def add_collision_constraints(self):
self.efc_D[n_con, i_b] = 1 / ti.max(diag, gs.EPS)
- @ti.kernel
- def add_equality_constraints(self):
- ti.loop_config(serialize=self._para_level < gs.PARA_LEVEL.PARTIAL)
- for i_b in range(self._B):
- fo... | genesis/engine/solvers/rigid/constraint_solver_decomp.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [FEATURE] Add equality joint constraint | 919 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/solvers/rigid/constraint_solver_decomp.py
PR Title: [FEATURE] Add equality joint constraint
BEFORE CODE: for j_d in range(i_d + 1, self._solver.n_dofs):
dot = gs.ti_float(0.0)
for k_d in range(i_d):
... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
qpos_start : None | array_like, optional
The start state. If None, the current state of the rigid entity will be used. Defaults to None.
timeout : float, optional
The maximum time (in seconds) allowed for the motion planning algorithm to find a solution. Defaults to 5.0.
... | No issues found. | qpos_start : None | array_like, optional
The start state. If None, the current state of the rigid entity will be used. Defaults to None.
timeout : float, optional
The maximum time (in seconds) allowed for the motion planning algorithm to find a solution. Defaults to 5.0.
... | @@ -363,7 +363,7 @@ def _load_MJCF(self, morph, surface):
)
def _load_URDF(self, morph, surface):
- l_infos, j_infos = uu.parse_urdf(morph, surface)
+ l_infos, j_infos, equalities = uu.parse_urdf(morph, surface)
for i_l in range(len(l_infos)):
l_info = l_inf... | genesis/engine/entities/rigid_entity/rigid_entity.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [FEATURE] Support mimic joint in URDF | 928 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | Review this pull request change.
Language: Python
File: genesis/engine/entities/rigid_entity/rigid_entity.py
PR Title: [FEATURE] Support mimic joint in URDF
BEFORE CODE: qpos_start : None | array_like, optional
The start state. If None, the current state of the rigid entity will be used. Default... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
quat = xyzw_to_wxyz(quat)
else:
gs.raise_exception(f"ndim expected to be 2 or 3, but got {T.ndim=}")
return trans, quat
else:
raise TypeError(f"Input must be a torch.Tensor or np.ndarray. Got: {type(T)}")
def trans_R_to_T(trans, R):
if isinstance(trans, torch.Te... | No issues found. | quat = xyzw_to_wxyz(quat)
else:
gs.raise_exception(f"ndim expected to be 2 or 3, but got {T.ndim=}")
return trans, quat
else:
raise TypeError(f"Input must be a torch.Tensor or np.ndarray. Got: {type(T)}")
def trans_R_to_T(trans, R):
if isinstance(trans, torch.Te... | @@ -21,6 +21,11 @@ def ti_transform_quat_by_quat(v, u):
return ti.Vector([w, x, y, z]).normalized()
+@ti.func
+def ti_quat_mul(u, v):
+ return ti_transform_quat_by_quat(v, u)
+
+
@ti.func
def ti_rotvec_to_quat(rotvec):
theta = rotvec.norm()
@@ -343,6 +348,18 @@ def get_face_norm(v0, v1, v2):
ret... | genesis/utils/geom.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [FEATURE] Add weld constraint | 948 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | Review this pull request change.
Language: Python
File: genesis/utils/geom.py
PR Title: [FEATURE] Add weld constraint
BEFORE CODE: quat = xyzw_to_wxyz(quat)
else:
gs.raise_exception(f"ndim expected to be 2 or 3, but got {T.ndim=}")
return trans, quat
else:
raise T... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
Whether to enable self collision within each entity. Defaults to False.
enable_adjacent_collision : bool, optional
Whether to enable collision between successive parent-child body pairs within each entity. Defaults to False.
max_collision_pairs : int, optional
Maximum number of collision... | No issues found. | Whether to enable self collision within each entity. Defaults to False.
enable_adjacent_collision : bool, optional
Whether to enable collision between successive parent-child body pairs within each entity. Defaults to False.
max_collision_pairs : int, optional
Maximum number of collision... | @@ -176,6 +176,8 @@ class RigidOptions(Options):
Velocity threshold for hibernation. Defaults to 1e-3.
hibernation_thresh_acc : float, optional
Acceleration threshold for hibernation. Defaults to 1e-2.
+ max_dynamic_constraints : int, optional
+ Maximum number of dynamic constraints (li... | genesis/options/solvers.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [FEATURE] Add suction cup | 961 | Genesis-Embodied-AI/Genesis | 26,702 | Python | YilingQiao | Review this pull request change.
Language: Python
File: genesis/options/solvers.py
PR Title: [FEATURE] Add suction cup
BEFORE CODE: Whether to enable self collision within each entity. Defaults to False.
enable_adjacent_collision : bool, optional
Whether to enable collision between successive pa... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
"runner_class_name": "OnPolicyRunner",
"num_steps_per_env": 100,
"save_interval": 100,
"empirical_normalization": None,
"seed": 1,
}
return train_cfg_dict
def get_cfgs():
env_cfg = {
"num_actions": 4,
# termination
"termination_if_roll_great... | No issues found. | "runner_class_name": "OnPolicyRunner",
"num_steps_per_env": 100,
"save_interval": 100,
"empirical_normalization": None,
"seed": 1,
}
return train_cfg_dict
def get_cfgs():
env_cfg = {
"num_actions": 4,
# termination
"termination_if_roll_great... | @@ -6,13 +6,13 @@
try:
try:
- if metadata.version("rsl-rl-lib"):
+ if metadata.version("rsl-rl"):
raise ImportError
except metadata.PackageNotFoundError:
- if metadata.version("rsl-rl") != "1.0.2":
+ if metadata.version("rsl-rl-lib") != "2.2.4":
raise I... | examples/drone/hover_train.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [FEATURE] DroneEnv compatible with rsl-rl-lib==2.2.4 | 1,036 | Genesis-Embodied-AI/Genesis | 26,702 | Python | KafuuChikai | Review this pull request change.
Language: Python
File: examples/drone/hover_train.py
PR Title: [FEATURE] DroneEnv compatible with rsl-rl-lib==2.2.4
BEFORE CODE: "runner_class_name": "OnPolicyRunner",
"num_steps_per_env": 100,
"save_interval": 100,
"empirical_normalization": None,
... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
[
self.sim.mpm_solver.boundary.lower,
self.sim.mpm_solver.boundary.upper,
]
),
wireframe=True,
color... | No issues found. | [
self.sim.mpm_solver.boundary.lower,
self.sim.mpm_solver.boundary.upper,
]
),
wireframe=True,
color... | @@ -648,6 +648,12 @@ def on_fem(self):
triangles_all[fem_entity.s_start : (fem_entity.s_start + fem_entity.n_surfaces)]
- fem_entity.v_start
)
+
+ # Select only vertices used in surface triangles, then reindex triangles against th... | genesis/vis/rasterizer_context.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [BUG FIX] Fix vertex index mismatch when adding tetrahedralized FEM entities. | 1,086 | Genesis-Embodied-AI/Genesis | 26,702 | Python | kosuke1701 | Review this pull request change.
Language: Python
File: genesis/vis/rasterizer_context.py
PR Title: [BUG FIX] Fix vertex index mismatch when adding tetrahedralized FEM entities.
BEFORE CODE: [
self.sim.mpm_solver.boundary.lower,
... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
if self.coacd_options is None:
self.coacd_options = CoacdOptions()
if isinstance(self.file, str):
file = os.path.abspath(self.file)
if not os.path.exists(file):
file = os.path.join(gs.utils.get_assets_dir(), self.file)
if not os.path.exi... | No issues found. | if self.coacd_options is None:
self.coacd_options = CoacdOptions()
if isinstance(self.file, str):
file = os.path.abspath(self.file)
if not os.path.exists(file):
file = os.path.join(gs.utils.get_assets_dir(), self.file)
if not os.path.exi... | @@ -16,6 +16,25 @@
"""
+class TetGenMixin(Options):
+ """
+ A mixin to introduce TetGen-related options into morph classes that support tetrahedralization using TetGen.
+ """
+
+ # FEM specific
+ order: int = 1
+
+ # Volumetric mesh entity
+ mindihedral: int = 10
+ minratio: float = 1.1
+ ... | genesis/options/morphs.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [FEATURE][BREAKING] Add maxvolume support for TetGen-based tetrahedralization, unify TetGen config handling, and add tests | 1,088 | Genesis-Embodied-AI/Genesis | 26,702 | Python | kosuke1701 | Review this pull request change.
Language: Python
File: genesis/options/morphs.py
PR Title: [FEATURE][BREAKING] Add maxvolume support for TetGen-based tetrahedralization, unify TetGen config handling, and add tests
BEFORE CODE: if self.coacd_options is None:
self.coacd_options = CoacdOptions()
... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
morph.scale * joint.limit.lower if joint.limit.lower is not None else -np.inf,
morph.scale * joint.limit.upper if joint.limit.upper is not None else np.inf,
]
]
)
j_info["dofs_stiffness"] = np.array([0.0])
... | No issues found. | morph.scale * joint.limit.lower if joint.limit.lower is not None else -np.inf,
morph.scale * joint.limit.upper if joint.limit.upper is not None else np.inf,
]
]
)
j_info["dofs_stiffness"] = np.array([0.0])
... | @@ -241,8 +241,8 @@ def parse_urdf(morph, surface):
j_info["dofs_limit"] = np.array(
[
[
- scale * joint.limit.lower if joint.limit.lower is not None else -np.inf,
- scale * joint.limit.upper if joint.limit.upper is not Non... | genesis/utils/urdf.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [BUG FIX] Fix scaling of poly-articulated robots. | 1,108 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | Review this pull request change.
Language: Python
File: genesis/utils/urdf.py
PR Title: [BUG FIX] Fix scaling of poly-articulated robots.
BEFORE CODE: morph.scale * joint.limit.lower if joint.limit.lower is not None else -np.inf,
morph.scale * joint.limit.upper if joi... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
init_simulators(gs_sim)
# Run the simulation for a few steps
for i in range(5000):
gs_sim.scene.step()
if i > 4900:
(gs_robot,) = gs_sim.entities
qvel = gs_robot.get_dofs_velocity().cpu()
assert_allclose(qvel, 0, atol=0.4)
qpos = gs_robot.get_dofs_po... | No issues found. | init_simulators(gs_sim)
# Run the simulation for a few steps
for i in range(5000):
gs_sim.scene.step()
if i > 4900:
(gs_robot,) = gs_sim.entities
qvel = gs_robot.get_dofs_velocity().cpu()
assert_allclose(qvel, 0, atol=0.4)
qpos = gs_robot.get_dofs_po... | @@ -1005,6 +1005,38 @@ def test_suction_cup(mode, show_viewer):
move_cube(use_suction=True, mode=mode, show_viewer=show_viewer)
+@pytest.mark.parametrize("backend", [gs.cpu, gs.gpu])
+def test_mass_mat(show_viewer, tol):
+ # Create and build the scene
+ scene = gs.Scene(
+ sim_options=gs.options.S... | tests/test_rigid_physics.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [FEATURE] get mass_mat API | 1,132 | Genesis-Embodied-AI/Genesis | 26,702 | Python | Kashu7100 | Review this pull request change.
Language: Python
File: tests/test_rigid_physics.py
PR Title: [FEATURE] get mass_mat API
BEFORE CODE: init_simulators(gs_sim)
# Run the simulation for a few steps
for i in range(5000):
gs_sim.scene.step()
if i > 4900:
(gs_robot,) = gs_sim.enti... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
Parameters
----------
force_field : gs.force_fields.ForceField
The force field to add to the scene.
Returns
-------
force_field : gs.force_fields.ForceField
The added force field.
"""
force_field.scene = self
self._sim._ad... | No issues found. |
Parameters
----------
force_field : gs.force_fields.ForceField
The force field to add to the scene.
Returns
-------
force_field : gs.force_fields.ForceField
The added force field.
"""
force_field.scene = self
self._sim._ad... | @@ -2,6 +2,11 @@
import torch
import genesis as gs
+from genesis.engine.entities.base_entity import Entity
+from genesis.engine.force_fields import ForceField
+from genesis.engine.materials.base import Material
+from genesis.options.morphs import Morph
+from genesis.options.surfaces import Surface
import genesis.u... | genesis/engine/scene.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [MISC] Add some initial mypy typing | 1,144 | Genesis-Embodied-AI/Genesis | 26,702 | Python | hughperkins | Review this pull request change.
Language: Python
File: genesis/engine/scene.py
PR Title: [MISC] Add some initial mypy typing
BEFORE CODE:
Parameters
----------
force_field : gs.force_fields.ForceField
The force field to add to the scene.
Returns
-------
... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
# Get more time to the robot to reach the last waypoint
for i in range(100):
scene.step()
# reach
qpos = franka.inverse_kinematics(
link=end_effector,
pos=np.array([0.65, 0.0, 0.13]),
quat=np.array([0, 1, 0, 0]),
)
franka.control_dofs_position(qpos[:-2], motors_... | No issues found. |
# Get more time to the robot to reach the last waypoint
for i in range(100):
scene.step()
# reach
qpos = franka.inverse_kinematics(
link=end_effector,
pos=np.array([0.65, 0.0, 0.13]),
quat=np.array([0, 1, 0, 0]),
)
franka.control_dofs_position(qpos[:-2], motors_... | @@ -804,37 +804,63 @@ def test_pd_control(show_viewer):
@pytest.mark.required
-def test_set_root_pose(show_viewer, tol):
+@pytest.mark.parametrize("relative", [False, True])
+def test_set_root_pose(relative, show_viewer, tol):
+ ROBOT_POS_ZERO = (0.0, 0.4, 0.1)
+ ROBOT_EULER_ZERO = (0.0, 0.0, 90.0)
+ CUBE... | tests/test_rigid_physics.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [BUG FIX] Fix urdf support. | 1,235 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | Review this pull request change.
Language: Python
File: tests/test_rigid_physics.py
PR Title: [BUG FIX] Fix urdf support.
BEFORE CODE:
# Get more time to the robot to reach the last waypoint
for i in range(100):
scene.step()
# reach
qpos = franka.inverse_kinematics(
link=end_effect... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
# DEBUG
# self._unif_map = {}
def _add_to_context(self):
if self._program_id is not None:
raise ValueError("Shader program already in context")
shader_ids = []
# Load vert shader
shader_ids.append(gl_shader_utils.compileShader(self._load(self.vertex_sha... | No issues found. |
# DEBUG
# self._unif_map = {}
def _add_to_context(self):
if self._program_id is not None:
raise ValueError("Shader program already in context")
shader_ids = []
# Load vert shader
shader_ids.append(gl_shader_utils.compileShader(self._load(self.vertex_sha... | @@ -221,14 +221,9 @@ def set_uniform(self, name, value, unsigned=False):
func1(loc, 1, GL_TRUE, value)
# Call correct uniform function
- elif isinstance(value, float):
- glUniform1f(loc, value)
- elif isinstance(value, int):
- if unsigned:
- ... | genesis/ext/pyrender/shader_program.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [BUG FIX] Various minor bug fixes. | 1,256 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | Review this pull request change.
Language: Python
File: genesis/ext/pyrender/shader_program.py
PR Title: [BUG FIX] Various minor bug fixes.
BEFORE CODE:
# DEBUG
# self._unif_map = {}
def _add_to_context(self):
if self._program_id is not None:
raise ValueError("Shader progra... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
)
actuator_kp = gu.default_dofs_kp(1)[0]
actuator_kv = gu.default_dofs_kv(1)[0]
elif gainprm[0] != -biasprm[1]:
# Doing our best to approximate the expected behavior: g0 * p_target + b1 * p_mes + b2 * v_mes
... | No issues found. | )
actuator_kp = gu.default_dofs_kp(1)[0]
actuator_kv = gu.default_dofs_kv(1)[0]
elif gainprm[0] != -biasprm[1]:
# Doing our best to approximate the expected behavior: g0 * p_target + b1 * p_mes + b2 * v_mes
... | @@ -20,7 +20,10 @@
from .misc import get_assets_dir, redirect_libc_stderr
-def build_model(xml, discard_visual, merge_fixed_links=False, links_to_keep=()):
+MIN_TIMECONST = np.finfo(np.double).eps
+
+
+def build_model(xml, discard_visual, default_armature=None, merge_fixed_links=False, links_to_keep=()):
if i... | genesis/utils/mjcf.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [BUG FIX] More robust MJCF / URDF parsing. | 1,262 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | Review this pull request change.
Language: Python
File: genesis/utils/mjcf.py
PR Title: [BUG FIX] More robust MJCF / URDF parsing.
BEFORE CODE: )
actuator_kp = gu.default_dofs_kp(1)[0]
actuator_kv = gu.default_dofs_kv(1)[0]
elif gainprm[0] ... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
i_ga, i_gb, i_b
)
if is_col:
# 1. Project the contact point on both geometries
# 2. Revert the effect of sm... | No issues found. | i_ga, i_gb, i_b
)
if is_col:
# 1. Project the contact point on both geometries
# 2. Revert the effect of sm... | @@ -302,72 +302,6 @@ def _func_is_geom_aabbs_overlap(self, i_ga, i_gb, i_b):
or (self._solver.geoms_state[i_ga, i_b].aabb_min >= self._solver.geoms_state[i_gb, i_b].aabb_max).any()
)
- @ti.func
- def _func_convex_geoms_overlap_ratio(self, i_ga, i_gb, i_b):
- # Check if one geom 'i_g... | genesis/engine/solvers/rigid/collider_decomp.py | 0 | Python | 1 | none | 16 | 51 | 51 | true | [MISC] Optimise benchmark duration. | 1,281 | Genesis-Embodied-AI/Genesis | 26,702 | Python | duburcqa | Review this pull request change.
Language: Python
File: genesis/engine/solvers/rigid/collider_decomp.py
PR Title: [MISC] Optimise benchmark duration.
BEFORE CODE: i_ga, i_gb, i_b
)
if is_col:... | No issues found. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.