torchrir 0.1.2__py3-none-any.whl → 0.1.4__py3-none-any.whl

torchrir/__init__.py

@@ -26,7 +26,12 @@ from .datasets import (
     load_wav_mono,
     save_wav,
 )
-from .scene_utils import binaural_mic_positions, clamp_positions, linear_trajectory, sample_positions
+from .scene_utils import (
+    binaural_mic_positions,
+    clamp_positions,
+    linear_trajectory,
+    sample_positions,
+)
 from .utils import (
     att2t_SabineEstimation,
     att2t_sabine_estimation,

torchrir/animation.py

@@ -104,15 +104,15 @@ def animate_scene_gif(
     mic_lines = []
     for _ in range(view_src_traj.shape[1]):
         if view_dim == 2:
-            line, = ax.plot([], [], color="tab:green", alpha=0.6)
+            (line,) = ax.plot([], [], color="tab:green", alpha=0.6)
         else:
-            line, = ax.plot([], [], [], color="tab:green", alpha=0.6)
+            (line,) = ax.plot([], [], [], color="tab:green", alpha=0.6)
         src_lines.append(line)
     for _ in range(view_mic_traj.shape[1]):
         if view_dim == 2:
-            line, = ax.plot([], [], color="tab:orange", alpha=0.6)
+            (line,) = ax.plot([], [], color="tab:orange", alpha=0.6)
         else:
-            line, = ax.plot([], [], [], color="tab:orange", alpha=0.6)
+            (line,) = ax.plot([], [], [], color="tab:orange", alpha=0.6)
         mic_lines.append(line)
 
     ax.legend(loc="best")
@@ -137,15 +137,15 @@ def animate_scene_gif(
                 xy = mic_frame[:, m_idx, :]
                 line.set_data(xy[:, 0], xy[:, 1])
         else:
-            src_scatter._offsets3d = (
-                src_pos_frame[:, 0],
-                src_pos_frame[:, 1],
-                src_pos_frame[:, 2],
+            setattr(
+                src_scatter,
+                "_offsets3d",
+                (src_pos_frame[:, 0], src_pos_frame[:, 1], src_pos_frame[:, 2]),
             )
-            mic_scatter._offsets3d = (
-                mic_pos_frame[:, 0],
-                mic_pos_frame[:, 1],
-                mic_pos_frame[:, 2],
+            setattr(
+                mic_scatter,
+                "_offsets3d",
+                (mic_pos_frame[:, 0], mic_pos_frame[:, 1], mic_pos_frame[:, 2]),
             )
             for s_idx, line in enumerate(src_lines):
                 xyz = src_frame[:, s_idx, :]
@@ -166,7 +166,10 @@ def animate_scene_gif(
             fps = frames / duration_s
         else:
             fps = 6.0
-    anim = animation.FuncAnimation(fig, _frame, frames=frames, interval=1000 / fps, blit=False)
-    anim.save(out_path, writer="pillow", fps=fps)
+    anim = animation.FuncAnimation(
+        fig, _frame, frames=frames, interval=1000 / fps, blit=False
+    )
+    fps_int = None if fps is None else max(1, int(round(fps)))
+    anim.save(out_path, writer="pillow", fps=fps_int)
     plt.close(fig)
     return out_path

torchrir/core.py

@@ -3,6 +3,7 @@ from __future__ import annotations
 """Core RIR simulation functions (static and dynamic)."""
 
 import math
+from collections.abc import Callable
 from typing import Optional, Tuple
 
 import torch
@@ -61,8 +62,8 @@ def simulate_rir(
 
     Example:
         >>> room = Room.shoebox(size=[6.0, 4.0, 3.0], fs=16000, beta=[0.9] * 6)
-        >>> sources = Source.positions([[1.0, 2.0, 1.5]])
-        >>> mics = MicrophoneArray.positions([[2.0, 2.0, 1.5]])
+        >>> sources = Source.from_positions([[1.0, 2.0, 1.5]])
+        >>> mics = MicrophoneArray.from_positions([[2.0, 2.0, 1.5]])
         >>> rir = simulate_rir(
         ...     room=room,
         ...     sources=sources,
@@ -90,9 +91,9 @@ def simulate_rir(
 
     if not isinstance(room, Room):
         raise TypeError("room must be a Room instance")
-    if nsample is None and tmax is None:
-        raise ValueError("nsample or tmax must be provided")
     if nsample is None:
+        if tmax is None:
+            raise ValueError("nsample or tmax must be provided")
         nsample = int(math.ceil(tmax * room.fs))
     if nsample <= 0:
         raise ValueError("nsample must be positive")
@@ -495,7 +496,11 @@ def _compute_image_contributions_batch(
     if mic_pattern != "omni":
         if mic_dir is None:
             raise ValueError("mic orientation required for non-omni directivity")
-        mic_dir = mic_dir[None, :, None, :] if mic_dir.ndim == 2 else mic_dir.view(1, 1, 1, -1)
+        mic_dir = (
+            mic_dir[None, :, None, :]
+            if mic_dir.ndim == 2
+            else mic_dir.view(1, 1, 1, -1)
+        )
         cos_theta = _cos_between(-vec, mic_dir)
         gain = gain * directivity_gain(mic_pattern, cos_theta)
 
@@ -542,9 +547,9 @@ def _accumulate_rir(
     if use_lut:
         sinc_lut = _get_sinc_lut(fdl, lut_gran, device=rir.device, dtype=dtype)
 
-    mic_offsets = (torch.arange(n_mic, device=rir.device, dtype=torch.int64) * nsample).view(
-        n_mic, 1, 1
-    )
+    mic_offsets = (
+        torch.arange(n_mic, device=rir.device, dtype=torch.int64) * nsample
+    ).view(n_mic, 1, 1)
     rir_flat = rir.view(-1)
 
     chunk_size = cfg.accumulate_chunk_size
@@ -559,7 +564,9 @@ def _accumulate_rir(
             x_off_frac = (1.0 - frac_m) * lut_gran
             lut_gran_off = torch.floor(x_off_frac).to(torch.int64)
             x_off = x_off_frac - lut_gran_off.to(dtype)
-            lut_pos = lut_gran_off[..., None] + (n[None, None, :].to(torch.int64) * lut_gran)
+            lut_pos = lut_gran_off[..., None] + (
+                n[None, None, :].to(torch.int64) * lut_gran
+            )
 
             s0 = torch.take(sinc_lut, lut_pos)
             s1 = torch.take(sinc_lut, lut_pos + 1)
@@ -618,9 +625,9 @@ def _accumulate_rir_batch_impl(
     if use_lut:
         sinc_lut = _get_sinc_lut(fdl, lut_gran, device=rir.device, dtype=sample.dtype)
 
-    sm_offsets = (torch.arange(n_sm, device=rir.device, dtype=torch.int64) * nsample).view(
-        n_sm, 1, 1
-    )
+    sm_offsets = (
+        torch.arange(n_sm, device=rir.device, dtype=torch.int64) * nsample
+    ).view(n_sm, 1, 1)
     rir_flat = rir.view(-1)
 
     n_img = idx0.shape[1]
@@ -634,7 +641,9 @@ def _accumulate_rir_batch_impl(
             x_off_frac = (1.0 - frac_m) * lut_gran
             lut_gran_off = torch.floor(x_off_frac).to(torch.int64)
             x_off = x_off_frac - lut_gran_off.to(sample.dtype)
-            lut_pos = lut_gran_off[..., None] + (n[None, None, :].to(torch.int64) * lut_gran)
+            lut_pos = lut_gran_off[..., None] + (
+                n[None, None, :].to(torch.int64) * lut_gran
+            )
 
             s0 = torch.take(sinc_lut, lut_pos)
             s1 = torch.take(sinc_lut, lut_pos + 1)
@@ -660,12 +669,13 @@ _SINC_LUT_CACHE: dict[tuple[int, int, str, torch.dtype], Tensor] = {}
 _FDL_GRID_CACHE: dict[tuple[int, str, torch.dtype], Tensor] = {}
 _FDL_OFFSETS_CACHE: dict[tuple[int, str], Tensor] = {}
 _FDL_WINDOW_CACHE: dict[tuple[int, str, torch.dtype], Tensor] = {}
-_ACCUM_BATCH_COMPILED: dict[tuple[str, torch.dtype, int, int, bool, int], callable] = {}
+_AccumFn = Callable[[Tensor, Tensor, Tensor], None]
+_ACCUM_BATCH_COMPILED: dict[tuple[str, torch.dtype, int, int, bool, int], _AccumFn] = {}
 
 
 def _get_accumulate_fn(
     cfg: SimulationConfig, device: torch.device, dtype: torch.dtype
-) -> callable:
+) -> _AccumFn:
     """Return an accumulation function with config-bound constants."""
     use_lut = cfg.use_lut and device.type != "mps"
     fdl = cfg.frac_delay_length
@@ -721,7 +731,9 @@ def _get_fdl_window(fdl: int, *, device: torch.device, dtype: torch.dtype) -> Te
     return cached
 
 
-def _get_sinc_lut(fdl: int, lut_gran: int, *, device: torch.device, dtype: torch.dtype) -> Tensor:
+def _get_sinc_lut(
+    fdl: int, lut_gran: int, *, device: torch.device, dtype: torch.dtype
+) -> Tensor:
     """Create a sinc lookup table for fractional delays."""
     key = (fdl, lut_gran, str(device), dtype)
     cached = _SINC_LUT_CACHE.get(key)
@@ -765,7 +777,12 @@ def _apply_diffuse_tail(
 
     gen = torch.Generator(device=rir.device)
     gen.manual_seed(0 if seed is None else seed)
-    noise = torch.randn(rir[..., tdiff_idx:].shape, device=rir.device, dtype=rir.dtype, generator=gen)
-    scale = torch.linalg.norm(rir[..., tdiff_idx - 1 : tdiff_idx], dim=-1, keepdim=True) + 1e-8
+    noise = torch.randn(
+        rir[..., tdiff_idx:].shape, device=rir.device, dtype=rir.dtype, generator=gen
+    )
+    scale = (
+        torch.linalg.norm(rir[..., tdiff_idx - 1 : tdiff_idx], dim=-1, keepdim=True)
+        + 1e-8
+    )
     rir[..., tdiff_idx:] = noise * decay * scale
     return rir

torchrir/datasets/cmu_arctic.py

@@ -44,6 +44,7 @@ def list_cmu_arctic_speakers() -> List[str]:
 @dataclass
 class CmuArcticSentence:
     """Sentence metadata from CMU ARCTIC."""
+
     utterance_id: str
     text: str
 
@@ -56,7 +57,9 @@ class CmuArcticDataset:
         >>> audio, fs = dataset.load_wav("arctic_a0001")
     """
 
-    def __init__(self, root: Path, speaker: str = "bdl", download: bool = False) -> None:
+    def __init__(
+        self, root: Path, speaker: str = "bdl", download: bool = False
+    ) -> None:
         """Initialize a CMU ARCTIC dataset handle.
 
         Args:

torchrir/datasets/template.py

@@ -34,7 +34,9 @@ class TemplateDataset(BaseDataset):
         protocol intact.
     """
 
-    def __init__(self, root: Path, speaker: str = "default", download: bool = False) -> None:
+    def __init__(
+        self, root: Path, speaker: str = "default", download: bool = False
+    ) -> None:
         self.root = Path(root)
         self.speaker = speaker
         if download:

torchrir/datasets/utils.py

@@ -10,7 +10,9 @@ import torch
 from .base import BaseDataset, SentenceLike
 
 
-def choose_speakers(dataset: BaseDataset, num_sources: int, rng: random.Random) -> List[str]:
+def choose_speakers(
+    dataset: BaseDataset, num_sources: int, rng: random.Random
+) -> List[str]:
     """Select unique speakers for the requested number of sources.
 
     Example:
@@ -89,4 +91,6 @@ def load_dataset_sources(
         info.append((speaker, utterance_ids))
 
     stacked = torch.stack(signals, dim=0)
+    if fs is None:
+        raise RuntimeError("no audio loaded from dataset sources")
     return stacked, int(fs), info

torchrir/dynamic.py

@@ -44,7 +44,9 @@ class DynamicConvolver:
             if self.hop is None:
                 raise ValueError("hop must be provided for hop mode")
             return _convolve_dynamic_hop(signal, rirs, self.hop)
-        return _convolve_dynamic_trajectory(signal, rirs, timestamps=self.timestamps, fs=self.fs)
+        return _convolve_dynamic_trajectory(
+            signal, rirs, timestamps=self.timestamps, fs=self.fs
+        )
 
 
 def _convolve_dynamic_hop(signal: Tensor, rirs: Tensor, hop: int) -> Tensor:

torchrir/plotting.py

@@ -92,7 +92,9 @@ def plot_scene_dynamic(
     return ax
 
 
-def _setup_axes(ax: Any | None, room: Room | Sequence[float] | Tensor) -> tuple[Any, Any]:
+def _setup_axes(
+    ax: Any | None, room: Room | Sequence[float] | Tensor
+) -> tuple[Any, Any]:
     """Create 2D/3D axes based on room dimension."""
     import matplotlib.pyplot as plt
 
@@ -131,8 +133,9 @@ def _draw_room_2d(ax: Any, size: Tensor) -> None:
     """Draw a 2D rectangular room."""
     import matplotlib.patches as patches
 
-    rect = patches.Rectangle((0.0, 0.0), size[0].item(), size[1].item(),
-                             fill=False, edgecolor="black")
+    rect = patches.Rectangle(
+        (0.0, 0.0), size[0].item(), size[1].item(), fill=False, edgecolor="black"
+    )
     ax.add_patch(rect)
     ax.set_xlim(0, size[0].item())
     ax.set_ylim(0, size[1].item())
@@ -186,7 +189,9 @@ def _draw_room_3d(ax: Any, size: Tensor) -> None:
     ax.set_zlabel("z")
 
 
-def _extract_positions(entity: Source | MicrophoneArray | Tensor | Sequence, ax: Any | None) -> Tensor:
+def _extract_positions(
+    entity: Source | MicrophoneArray | Tensor | Sequence, ax: Any | None
+) -> Tensor:
     """Extract positions from Source/MicrophoneArray or raw tensor."""
     if isinstance(entity, (Source, MicrophoneArray)):
         pos = entity.positions
@@ -211,7 +216,9 @@ def _scatter_positions(
         return
     dim = positions.shape[1]
     if dim == 2:
-        ax.scatter(positions[:, 0], positions[:, 1], label=label, marker=marker, color=color)
+        ax.scatter(
+            positions[:, 0], positions[:, 1], label=label, marker=marker, color=color
+        )
     else:
         ax.scatter(
             positions[:, 0],
@@ -300,4 +307,4 @@ def _is_moving(traj: Tensor, positions: Tensor, *, tol: float = 1e-6) -> bool:
     if traj.numel() == 0:
         return False
     pos0 = positions.unsqueeze(0).expand_as(traj)
-    return torch.any(torch.linalg.norm(traj - pos0, dim=-1) > tol).item()
+    return bool(torch.any(torch.linalg.norm(traj - pos0, dim=-1) > tol).item())

torchrir/plotting_utils.py

@@ -127,7 +127,10 @@ def _positions_to_cpu(entity: torch.Tensor | object) -> torch.Tensor:
     return pos
 
 
-def _traj_steps(src_traj: Optional[torch.Tensor | Sequence], mic_traj: Optional[torch.Tensor | Sequence]) -> int:
+def _traj_steps(
+    src_traj: Optional[torch.Tensor | Sequence],
+    mic_traj: Optional[torch.Tensor | Sequence],
+) -> int:
     """Infer the number of trajectory steps."""
     if src_traj is not None:
         return int(_to_cpu(src_traj).shape[0])

torchrir/room.py

@@ -65,7 +65,7 @@ class Source:
     """Source container with positions and optional orientation.
 
     Example:
-        >>> sources = Source.positions([[1.0, 2.0, 1.5]])
+        >>> sources = Source.from_positions([[1.0, 2.0, 1.5]])
     """
 
     positions: Tensor
@@ -82,24 +82,6 @@ class Source:
         """Return a new Source with updated fields."""
         return replace(self, **kwargs)
 
-    @classmethod
-    def positions(
-        cls,
-        positions: Sequence[Sequence[float]] | Tensor,
-        *,
-        orientation: Optional[Sequence[float] | Tensor] = None,
-        device: Optional[torch.device | str] = None,
-        dtype: Optional[torch.dtype] = None,
-    ) -> "Source":
-        """Construct a Source from positions.
-
-        Example:
-            >>> sources = Source.positions([[1.0, 2.0, 1.5]])
-        """
-        return cls.from_positions(
-            positions, orientation=orientation, device=device, dtype=dtype
-        )
-
     @classmethod
     def from_positions(
         cls,
@@ -122,7 +104,7 @@ class MicrophoneArray:
     """Microphone array container.
 
     Example:
-        >>> mics = MicrophoneArray.positions([[2.0, 2.0, 1.5]])
+        >>> mics = MicrophoneArray.from_positions([[2.0, 2.0, 1.5]])
     """
 
     positions: Tensor
@@ -139,24 +121,6 @@ class MicrophoneArray:
         """Return a new MicrophoneArray with updated fields."""
         return replace(self, **kwargs)
 
-    @classmethod
-    def positions(
-        cls,
-        positions: Sequence[Sequence[float]] | Tensor,
-        *,
-        orientation: Optional[Sequence[float] | Tensor] = None,
-        device: Optional[torch.device | str] = None,
-        dtype: Optional[torch.dtype] = None,
-    ) -> "MicrophoneArray":
-        """Construct a MicrophoneArray from positions.
-
-        Example:
-            >>> mics = MicrophoneArray.positions([[2.0, 2.0, 1.5]])
-        """
-        return cls.from_positions(
-            positions, orientation=orientation, device=device, dtype=dtype
-        )
-
     @classmethod
     def from_positions(
         cls,

torchrir/scene_utils.py

@@ -32,7 +32,9 @@ def sample_positions(
     return torch.tensor(coords, dtype=torch.float32)
 
 
-def linear_trajectory(start: torch.Tensor, end: torch.Tensor, steps: int) -> torch.Tensor:
+def linear_trajectory(
+    start: torch.Tensor, end: torch.Tensor, steps: int
+) -> torch.Tensor:
     """Create a linear trajectory between start and end.
 
     Example:
@@ -58,7 +60,9 @@ def binaural_mic_positions(center: torch.Tensor, offset: float = 0.08) -> torch.
     return torch.stack([left, right], dim=0)
 
 
-def clamp_positions(positions: torch.Tensor, room_size: torch.Tensor, margin: float = 0.1) -> torch.Tensor:
+def clamp_positions(
+    positions: torch.Tensor, room_size: torch.Tensor, margin: float = 0.1
+) -> torch.Tensor:
     """Clamp positions to remain inside the room with a margin.
 
     Example:

torchrir/signal.py

@@ -117,9 +117,9 @@ def _convolve_dynamic_rir_trajectory(
     else:
         step_fs = n_samples / t_steps
         ts_dtype = torch.float32 if signal.device.type == "mps" else torch.float64
-        w_ini = (torch.arange(t_steps, device=signal.device, dtype=ts_dtype) * step_fs).to(
-            torch.long
-        )
+        w_ini = (
+            torch.arange(t_steps, device=signal.device, dtype=ts_dtype) * step_fs
+        ).to(torch.long)
 
     w_ini = torch.cat(
         [w_ini, torch.tensor([n_samples], device=signal.device, dtype=torch.long)]
@@ -132,14 +132,18 @@ def _convolve_dynamic_rir_trajectory(
         )
 
     max_len = int(w_len.max().item())
-    segments = torch.zeros((t_steps, n_src, max_len), dtype=signal.dtype, device=signal.device)
+    segments = torch.zeros(
+        (t_steps, n_src, max_len), dtype=signal.dtype, device=signal.device
+    )
     for t in range(t_steps):
         start = int(w_ini[t].item())
         end = int(w_ini[t + 1].item())
         if end > start:
             segments[t, :, : end - start] = signal[:, start:end]
 
-    out = torch.zeros((n_mic, n_samples + rir_len - 1), dtype=signal.dtype, device=signal.device)
+    out = torch.zeros(
+        (n_mic, n_samples + rir_len - 1), dtype=signal.dtype, device=signal.device
+    )
 
     for t in range(t_steps):
         seg_len = int(w_len[t].item())
@@ -166,7 +170,9 @@ def _convolve_dynamic_rir_trajectory_batched(
     """GPU-friendly batched trajectory convolution using FFT."""
     n_samples = signal.shape[1]
     t_steps, n_src, n_mic, rir_len = rirs.shape
-    out = torch.zeros((n_mic, n_samples + rir_len - 1), dtype=signal.dtype, device=signal.device)
+    out = torch.zeros(
+        (n_mic, n_samples + rir_len - 1), dtype=signal.dtype, device=signal.device
+    )
 
     for t0 in range(0, t_steps, chunk_size):
         t1 = min(t0 + chunk_size, t_steps)
@@ -174,7 +180,9 @@ def _convolve_dynamic_rir_trajectory_batched(
         max_len = int(lengths.max().item())
         if max_len == 0:
             continue
-        segments = torch.zeros((t1 - t0, n_src, max_len), dtype=signal.dtype, device=signal.device)
+        segments = torch.zeros(
+            (t1 - t0, n_src, max_len), dtype=signal.dtype, device=signal.device
+        )
         for idx, t in enumerate(range(t0, t1)):
             start = int(w_ini[t].item())
             end = int(w_ini[t + 1].item())
@@ -190,7 +198,9 @@ def _convolve_dynamic_rir_trajectory_batched(
             dtype=signal.dtype,
             device=signal.device,
         )
-        conv = torch.fft.irfft(seg_f[:, :, None, :] * rir_f, n=fft_len, dim=-1, out=conv_out)
+        conv = torch.fft.irfft(
+            seg_f[:, :, None, :] * rir_f, n=fft_len, dim=-1, out=conv_out
+        )
         conv = conv[..., :conv_len]
         conv_sum = conv.sum(dim=1)
 
@@ -199,7 +209,9 @@ def _convolve_dynamic_rir_trajectory_batched(
             if seg_len == 0:
                 continue
             start = int(w_ini[t].item())
-            out[:, start : start + seg_len + rir_len - 1] += conv_sum[idx, :, : seg_len + rir_len - 1]
+            out[:, start : start + seg_len + rir_len - 1] += conv_sum[
+                idx, :, : seg_len + rir_len - 1
+            ]
 
     return out.squeeze(0) if n_mic == 1 else out
 
@@ -221,7 +233,9 @@ def _ensure_static_rirs(rirs: Tensor) -> Tensor:
         return rirs.view(1, rirs.shape[0], rirs.shape[1])
     if rirs.ndim == 3:
         return rirs
-    raise ValueError("rirs must have shape (rir_len,), (n_mic, rir_len), or (n_src, n_mic, rir_len)")
+    raise ValueError(
+        "rirs must have shape (rir_len,), (n_mic, rir_len), or (n_src, n_mic, rir_len)"
+    )
 
 
 def _ensure_dynamic_rirs(rirs: Tensor, signal: Tensor) -> Tensor:

torchrir/simulators.py

@@ -18,7 +18,9 @@ from .scene import Scene
 class RIRSimulator(Protocol):
     """Strategy interface for RIR simulation backends."""
 
-    def simulate(self, scene: Scene, config: SimulationConfig | None = None) -> RIRResult:
+    def simulate(
+        self, scene: Scene, config: SimulationConfig | None = None
+    ) -> RIRResult:
         """Run a simulation and return the result."""
 
 
@@ -30,7 +32,9 @@ class ISMSimulator:
         >>> result = ISMSimulator().simulate(scene, config)
     """
 
-    def simulate(self, scene: Scene, config: SimulationConfig | None = None) -> RIRResult:
+    def simulate(
+        self, scene: Scene, config: SimulationConfig | None = None
+    ) -> RIRResult:
         scene.validate()
         cfg = config or default_config()
         if scene.is_dynamic():
@@ -71,7 +75,9 @@ class RayTracingSimulator:
         reuse Scene/SimulationConfig for inputs and keep output shape parity.
     """
 
-    def simulate(self, scene: Scene, config: SimulationConfig | None = None) -> RIRResult:
+    def simulate(
+        self, scene: Scene, config: SimulationConfig | None = None
+    ) -> RIRResult:
         raise NotImplementedError("RayTracingSimulator is not implemented yet")
 
 
@@ -86,5 +92,7 @@ class FDTDSimulator:
         RIRResult with the same metadata contract as ISM.
     """
 
-    def simulate(self, scene: Scene, config: SimulationConfig | None = None) -> RIRResult:
+    def simulate(
+        self, scene: Scene, config: SimulationConfig | None = None
+    ) -> RIRResult:
         raise NotImplementedError("FDTDSimulator is not implemented yet")

torchrir/utils.py

@@ -15,7 +15,7 @@ _DEF_SPEED_OF_SOUND = 343.0
 
 
 def as_tensor(
-    value: Tensor | Iterable[float] | float | int,
+    value: Tensor | Iterable[float] | Iterable[Iterable[float]] | float | int,
     *,
     device: Optional[torch.device | str] = None,
     dtype: Optional[torch.dtype] = None,

torchrir-0.1.4.dist-info/METADATA

@@ -0,0 +1,70 @@
+Metadata-Version: 2.4
+Name: torchrir
+Version: 0.1.4
+Summary: PyTorch-based room impulse response (RIR) simulation toolkit for static and dynamic scenes.
+Project-URL: Repository, https://github.com/taishi-n/torchrir
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: NOTICE
+Requires-Dist: numpy>=2.2.6
+Requires-Dist: torch>=2.10.0
+Dynamic: license-file
+
+# TorchRIR
+
+PyTorch-based room impulse response (RIR) simulation toolkit focused on a clean, modern API with GPU support.
+This project has been substantially assisted by AI using Codex.
+
+## Installation
+```bash
+pip install torchrir
+```
+
+## Examples
+- `examples/static.py`: fixed sources/mics with binaural output.  
+  `uv run python examples/static.py --plot`
+- `examples/dynamic_src.py`: moving sources, fixed mics.  
+  `uv run python examples/dynamic_src.py --plot`
+- `examples/dynamic_mic.py`: fixed sources, moving mics.  
+  `uv run python examples/dynamic_mic.py --plot`
+- `examples/cli.py`: unified CLI for static/dynamic scenes, JSON/YAML configs.  
+  `uv run python examples/cli.py --mode static --plot`
+- `examples/cmu_arctic_dynamic_dataset.py`: small dynamic dataset generator (fixed room/mics, randomized source motion).  
+  `uv run python examples/cmu_arctic_dynamic_dataset.py --num-scenes 4 --num-sources 2`
+- `examples/benchmark_device.py`: CPU/GPU benchmark for RIR simulation.  
+  `uv run python examples/benchmark_device.py --dynamic`
+
+## Core API Overview
+- Geometry: `Room`, `Source`, `MicrophoneArray`
+- Static RIR: `simulate_rir`
+- Dynamic RIR: `simulate_dynamic_rir`
+- Dynamic convolution: `DynamicConvolver`
+- Metadata export: `build_metadata`, `save_metadata_json`
+
+```python
+from torchrir import DynamicConvolver, MicrophoneArray, Room, Source, simulate_rir
+
+room = Room.shoebox(size=[6.0, 4.0, 3.0], fs=16000, beta=[0.9] * 6)
+sources = Source.from_positions([[1.0, 2.0, 1.5]])
+mics = MicrophoneArray.from_positions([[2.0, 2.0, 1.5]])
+
+rir = simulate_rir(room=room, sources=sources, mics=mics, max_order=6, tmax=0.3)
+# For dynamic scenes, compute rirs with simulate_dynamic_rir and convolve:
+# y = DynamicConvolver(mode="trajectory").convolve(signal, rirs)
+```
+
+For detailed documentation, see the docs under `docs/` and Read the Docs.
+
+## Future Work
+- Ray tracing backend: implement `RayTracingSimulator` with frequency-dependent absorption/scattering.
+- CUDA-native acceleration: introduce dedicated CUDA kernels for large-scale RIR generation.
+- Dataset expansion: add additional dataset integrations beyond CMU ARCTIC (see `TemplateDataset`).
+- Add regression tests comparing generated RIRs against gpuRIR outputs.
+
+## Related Libraries
+- [gpuRIR](https://github.com/DavidDiazGuerra/gpuRIR)
+- [Cross3D](https://github.com/DavidDiazGuerra/Cross3D)
+- [pyroomacoustics](https://github.com/LCAV/pyroomacoustics)
+- [das-generator](https://github.com/ehabets/das-generator)
+- [rir-generator](https://github.com/audiolabs/rir-generator)

torchrir-0.1.4.dist-info/RECORD

@@ -0,0 +1,28 @@
+torchrir/__init__.py,sha256=urydbUWuUHPBqmy-9QBaQg8eFGznRamkSjLmPiNvBo0,2383
+torchrir/animation.py,sha256=x3Y-BLz3J6DQNmoDIjbMEgGfng2yavJFLyQEmRCSpQU,6391
+torchrir/config.py,sha256=PsZdDIS3p4jepeNSHyd69aSD9QlOEdpG9v1SAXlZ_Fg,2295
+torchrir/core.py,sha256=VdljYoCoQoZqD8aYJRnuHEb7uORQjyQysVc8K3RGuao,26826
+torchrir/directivity.py,sha256=v_t37YgeXF_IYzbnrk0TCs1npb_0yKR7zHiG8XV3V4w,1259
+torchrir/dynamic.py,sha256=01JHMxhORdcz93J-YaMIeSLo7k2tHrZke8llPHHXwZg,2153
+torchrir/logging_utils.py,sha256=s4jDSSDoHT0HKeplDUpGMsdeBij4eibLSpaaAPzkB68,2146
+torchrir/metadata.py,sha256=cwoXrr_yE2bQRUPnJe6p7POMPCWa9_oabCtp--WqBE8,6958
+torchrir/plotting.py,sha256=TM1LxPitZq5KXdNe1GfUCOnzFOzerGhWIFblzIz142A,8170
+torchrir/plotting_utils.py,sha256=Kg3TCLqEq_lxVQkYHI6vyz_6oG3Ic_Z8H9gZN-39QeI,5180
+torchrir/results.py,sha256=-HczEfr2u91BNb1xbrIGKCj0G3yzy7l_fmUMUeKbGRw,614
+torchrir/room.py,sha256=zFnEzw0Rr1NP9IUc3iNTInyoq6t3X-0yOyUtDnsLSPk,4325
+torchrir/scene.py,sha256=GuHuCspakAUOT81_ArTqaZbmBX0ApoJuCKTaZ21wGis,2435
+torchrir/scene_utils.py,sha256=2La5dtjxYdINX315VXRRJMJK9oaR2rY0xHmDLjZma8M,2140
+torchrir/signal.py,sha256=M0BpKDBqrfOmCHIJ_dvl-C3uKdFpXLDqtSIU115jsME,8383
+torchrir/simulators.py,sha256=NCl8Ptv2TGdBpNLwAb3nigT77On-BLIANtc2ivgKasw,3131
+torchrir/utils.py,sha256=2oE-JzAtkW5qdRds2Y5R5lbSyNZl_9piFXd6xOLzjxM,10680
+torchrir/datasets/__init__.py,sha256=3T55F3fjjRR3j618ubRkMlZnQTxvXaxioFMhygxm7oQ,601
+torchrir/datasets/base.py,sha256=mCHLtGOOaD1II1alJpP6ipzkz87l-rh19NgfeLnJbDU,720
+torchrir/datasets/cmu_arctic.py,sha256=DrOcawHvOEUnFJRw4qZgwuK1jbL2oQ-Vz_zNodYtpjE,7049
+torchrir/datasets/template.py,sha256=pHAKj5E7Gehfk9pqdTsFQjiDV1OK3hSZJIbYutd-E4c,2090
+torchrir/datasets/utils.py,sha256=TUfdt_XSB71ztCfzq_gCNrbvPh0Y-O5gkyxUnHWYID0,3227
+torchrir-0.1.4.dist-info/licenses/LICENSE,sha256=5vS_7WTsMEw_QQHEPQ_WCwovJXEgmxoEwcwOI-9VbXI,10766
+torchrir-0.1.4.dist-info/licenses/NOTICE,sha256=SRs_q-ZqoVF9_YuuedZOvVBk01jV7YQAeF8rRvlRg0s,118
+torchrir-0.1.4.dist-info/METADATA,sha256=HNSQV3uXeRYfX9eDb7ZllGAMvWCdtzq9Rn-q0kokkL4,2964
+torchrir-0.1.4.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+torchrir-0.1.4.dist-info/top_level.txt,sha256=aIFwntowJjvm7rZk480HymC3ipDo1g-9hEbNY1wF-Oo,9
+torchrir-0.1.4.dist-info/RECORD,,

torchrir-0.1.2.dist-info/METADATA

@@ -1,271 +0,0 @@
-Metadata-Version: 2.4
-Name: torchrir
-Version: 0.1.2
-Summary: PyTorch-based room impulse response (RIR) simulation toolkit for static and dynamic scenes.
-Project-URL: Repository, https://github.com/taishi-n/torchrir
-Requires-Python: >=3.10
-Description-Content-Type: text/markdown
-License-File: LICENSE
-License-File: NOTICE
-Requires-Dist: numpy>=2.2.6
-Requires-Dist: torch>=2.10.0
-Dynamic: license-file
-
-# TorchRIR
-
-PyTorch-based room impulse response (RIR) simulation toolkit focused on a clean, modern API with GPU support.
-This project has been substantially assisted by AI using Codex.
-
-## License
-Apache-2.0. See `LICENSE` and `NOTICE`.
-
-## Installation
-```bash
-pip install torchrir
-```
-
-## Current Capabilities
-- ISM-based static and dynamic RIR simulation (2D/3D shoebox rooms).
-- Directivity patterns: `omni`, `cardioid`, `hypercardioid`, `subcardioid`, `bidir` with orientation handling.
-- Acoustic parameters: `beta` or `t60` (Sabine), optional diffuse tail via `tdiff`.
-- Dynamic convolution via `DynamicConvolver` (`trajectory` or `hop` modes).
-- GPU acceleration for ISM accumulation (CUDA/MPS; MPS disables LUT).
-- Dataset utilities with CMU ARCTIC support and example pipelines.
-- Plotting utilities for static and dynamic scenes.
-- Metadata export helpers for time axis, DOA, and array attributes (JSON-ready).
-- Unified CLI with JSON/YAML config and deterministic flag support.
-
-## Example Usage
-```bash
-# CMU ARCTIC + static RIR (fixed sources/mics)
-uv run python examples/static.py --plot
-
-# Dynamic RIR demos
-uv run python examples/dynamic_mic.py --plot
-uv run python examples/dynamic_src.py --plot
-uv run python examples/dynamic_mic.py --gif
-uv run python examples/dynamic_src.py --gif
-
-# Unified CLI
-uv run python examples/cli.py --mode static --plot
-uv run python examples/cli.py --mode dynamic_mic --plot
-uv run python examples/cli.py --mode dynamic_src --plot
-uv run python examples/cli.py --mode dynamic_mic --gif
-uv run python examples/dynamic_mic.py --gif --gif-fps 12
-
-# Config + deterministic
-uv run python examples/cli.py --mode static --deterministic --seed 123 --config-out outputs/cli.json
-uv run python examples/cli.py --config-in outputs/cli.json
-```
-GIF FPS is auto-derived from signal duration and RIR steps unless overridden with `--gif-fps`.
-For 3D rooms, an additional `*_3d.gif` is saved.
-YAML configs are supported when `PyYAML` is installed.
-```bash
-# YAML config
-uv run python examples/cli.py --mode static --config-out outputs/cli.yaml
-uv run python examples/cli.py --config-in outputs/cli.yaml
-```
-`examples/cli_example.yaml` provides a ready-to-use template.
-Examples also save `*_metadata.json` alongside audio outputs.
-
-```python
-from torchrir import DynamicConvolver, MicrophoneArray, Room, Source, simulate_rir
-
-room = Room.shoebox(size=[6.0, 4.0, 3.0], fs=16000, beta=[0.9] * 6)
-sources = Source.positions([[1.0, 2.0, 1.5]])
-mics = MicrophoneArray.positions([[2.0, 2.0, 1.5]])
-
-rir = simulate_rir(
-    room=room,
-    sources=sources,
-    mics=mics,
-    max_order=6,
-    tmax=0.3,
-    device="auto",
-)
-```
-
-```python
-from torchrir import DynamicConvolver
-
-# Trajectory-mode dynamic convolution
-y = DynamicConvolver(mode="trajectory").convolve(signal, rirs)
-
-# Hop-mode dynamic convolution
-y = DynamicConvolver(mode="hop", hop=1024).convolve(signal, rirs)
-```
-Dynamic convolution is exposed via `DynamicConvolver` only (no legacy function wrappers).
-
-## Limitations and Potential Errors
-- Ray tracing and FDTD simulators are placeholders and raise `NotImplementedError`.
-- `TemplateDataset` methods are not implemented and will raise `NotImplementedError`.
-- `simulate_rir`/`simulate_dynamic_rir` require `max_order` (or `SimulationConfig.max_order`) and either `nsample` or `tmax`.
-- Non-`omni` directivity requires orientation; mismatched shapes raise `ValueError`.
-- `beta` must have 4 (2D) or 6 (3D) elements; invalid sizes raise `ValueError`.
-- `simulate_dynamic_rir` requires `src_traj` and `mic_traj` to have matching time steps.
-- Dynamic simulation currently loops per time step; very long trajectories can be slow.
-- MPS disables the sinc LUT path (falls back to direct sinc), which can be slower and slightly different numerically.
-- Deterministic mode is best-effort; some backends may still be non-deterministic.
-- YAML configs require `PyYAML`; otherwise a `ModuleNotFoundError` is raised.
-- CMU ARCTIC downloads require network access.
-- GIF animation output requires Pillow (via matplotlib animation writer).
-
-### Dataset-agnostic utilities
-```python
-from torchrir import (
-    CmuArcticDataset,
-    binaural_mic_positions,
-    clamp_positions,
-    load_dataset_sources,
-    sample_positions,
-)
-
-def dataset_factory(speaker: str | None):
-    spk = speaker or "bdl"
-    return CmuArcticDataset("datasets/cmu_arctic", speaker=spk, download=True)
-
-signals, fs, info = load_dataset_sources(
-    dataset_factory=dataset_factory,
-    num_sources=2,
-    duration_s=10.0,
-    rng=random.Random(0),
-)
-```
-
-### Dataset template (for future extension)
-`TemplateDataset` provides a minimal stub to implement new datasets later.
-
-### Logging
-```python
-from torchrir import LoggingConfig, get_logger, setup_logging
-
-setup_logging(LoggingConfig(level="INFO"))
-logger = get_logger("examples")
-logger.info("running torchrir example")
-```
-
-### Scene container
-```python
-from torchrir import Scene
-
-scene = Scene(room=room, sources=sources, mics=mics, src_traj=src_traj, mic_traj=mic_traj)
-scene.validate()
-```
-
-### Immutable geometry helpers
-`Room`, `Source`, and `MicrophoneArray` are immutable; use `.replace()` to update fields.
-
-### Result container
-```python
-from torchrir import RIRResult
-
-result = RIRResult(rirs=rirs, scene=scene, config=config)
-```
-
-### Simulation strategies
-```python
-from torchrir import ISMSimulator
-
-sim = ISMSimulator()
-result = sim.simulate(scene, config)
-```
-
-## Device Selection
-- `device="cpu"`: CPU execution
-- `device="cuda"`: NVIDIA GPU (CUDA) if available, otherwise fallback to CPU
-- `device="mps"`: Apple Silicon GPU via Metal (MPS) if available, otherwise fallback to CPU
-- `device="auto"`: prefer CUDA → MPS → CPU
-
-```python
-from torchrir import DeviceSpec
-
-device, dtype = DeviceSpec(device="auto").resolve()
-```
-
-## References
-- [gpuRIR](https://github.com/DavidDiazGuerra/gpuRIR)
-- [Cross3D](https://github.com/DavidDiazGuerra/Cross3D)
-- [pyroomacoustics](https://github.com/LCAV/pyroomacoustics)
-- [das-generator](https://github.com/ehabets/das-generator)
-- [rir-generator](https://github.com/audiolabs/rir-generator)
-
-## Specification (Current)
-### Purpose
-- Provide room impulse response (RIR) simulation on PyTorch with CPU/CUDA/MPS support.
-- Support static and dynamic scenes with a maintainable, modern API.
-
-### Room Model
-- Shoebox (rectangular) room model.
-- 2D or 3D.
-- Image Source Method (ISM) implementation.
-
-### Inputs
-#### Scene Geometry
-- Room size: `[Lx, Ly, Lz]` (2D uses `[Lx, Ly]`).
-- Source positions: `(n_src, dim)`.
-- Microphone positions: `(n_mic, dim)`.
-- Reflection order: `max_order`.
-
-#### Acoustic Parameters
-- Sample rate: `fs`.
-- Speed of sound: `c` (default 343.0 m/s).
-- Wall reflection coefficients: `beta` (4 faces for 2D, 6 for 3D) or `t60` (Sabine).
-
-#### Output Length
-- Specify `nsample` (samples) or `tmax` (seconds).
-
-#### Directivity
-- Patterns: `omni`, `cardioid`, `hypercardioid`, `subcardioid`, `bidir`.
-- Orientation specified by vector or angles.
-
-#### Configuration
-- `SimulationConfig` controls algorithm settings (e.g., max_order, tmax, directivity, device, seed, fractional delay length, LUT, chunk sizes, compile path).
-- Passed explicitly via `simulate_rir(..., config=...)` or `simulate_dynamic_rir(..., config=...)`.
-
-### Outputs
-- Static RIR shape: `(n_src, n_mic, nsample)`.
-- Dynamic RIR shape: `(T, n_src, n_mic, nsample)`.
-- Preserves dtype/device.
-
-### Core APIs
-#### Static RIR
-```python
-room = Room.shoebox(size=[6.0, 4.0, 3.0], fs=16000, beta=[0.9] * 6)
-sources = Source.positions([[1.0, 2.0, 1.5], [4.5, 1.0, 1.2]])
-mics = MicrophoneArray.positions([[2.0, 2.0, 1.5], [3.0, 2.0, 1.5]])
-
-rir = simulate_rir(
-    room=room,
-    sources=sources,
-    mics=mics,
-    max_order=8,
-    tmax=0.4,
-    directivity="omni",
-    device="auto",
-)
-```
-
-#### Dynamic RIRs + Convolution
-```python
-rirs = simulate_dynamic_rir(
-    room=room,
-    src_traj=src_traj,   # (T, n_src, dim)
-    mic_traj=mic_traj,   # (T, n_mic, dim)
-    max_order=8,
-    tmax=0.4,
-    device="auto",
-)
-
-y = DynamicConvolver(mode="trajectory").convolve(signal, rirs)
-```
-
-### Device Control
-- `device="cpu"`, `"cuda"`, `"mps"`, or `"auto"`; resolves with fallback to CPU.
-
-## Future Work
-- Ray tracing backend: implement `RayTracingSimulator` with frequency-dependent absorption/scattering.
-- FDTD backend: implement `FDTDSimulator` with configurable grid resolution and boundary conditions.
-- Dataset expansion: add additional dataset integrations beyond CMU ARCTIC (see `TemplateDataset`).
-- Enhanced acoustics: frequency-dependent absorption and more advanced diffuse tail models.
-- Add microphone and source directivity models similar to gpuRIR/pyroomacoustics.
-- Add regression tests comparing generated RIRs against gpuRIR outputs.

torchrir-0.1.2.dist-info/RECORD

@@ -1,28 +0,0 @@
-torchrir/__init__.py,sha256=gUaLrXkWTqDSHCvU7JEB39Ynh-mWAWQGmnbV7Zghi2w,2362
-torchrir/animation.py,sha256=bQ-yWRV7oCzaFU5I1J-VdsiQAWqU1--R6vH6asxQhRw,6280
-torchrir/config.py,sha256=PsZdDIS3p4jepeNSHyd69aSD9QlOEdpG9v1SAXlZ_Fg,2295
-torchrir/core.py,sha256=1Ups0_os9ArbSA3RIPeS8ftGSygEfwCsoc28u6xy6tg,26587
-torchrir/directivity.py,sha256=v_t37YgeXF_IYzbnrk0TCs1npb_0yKR7zHiG8XV3V4w,1259
-torchrir/dynamic.py,sha256=j8W-2Xwak_v4GQ-RetzlWXyHVUmbTFs1H7dEIwYYXlA,2131
-torchrir/logging_utils.py,sha256=s4jDSSDoHT0HKeplDUpGMsdeBij4eibLSpaaAPzkB68,2146
-torchrir/metadata.py,sha256=cwoXrr_yE2bQRUPnJe6p7POMPCWa9_oabCtp--WqBE8,6958
-torchrir/plotting.py,sha256=ynimJC10QRTMVIC5vH3w1bMEN3e-0CyPHICtZzWu3XE,8145
-torchrir/plotting_utils.py,sha256=uGceTrwWmJhZ_V1FnJNp30vE13hHbXDkY3X0aAxDhvc,5169
-torchrir/results.py,sha256=-HczEfr2u91BNb1xbrIGKCj0G3yzy7l_fmUMUeKbGRw,614
-torchrir/room.py,sha256=BnqA4FY_k_RQV9PvUAX8fr15yIP-PBIrzkmAiwLPZHU,5449
-torchrir/scene.py,sha256=GuHuCspakAUOT81_ArTqaZbmBX0ApoJuCKTaZ21wGis,2435
-torchrir/scene_utils.py,sha256=Kkj1XL1Xtpf_2bk7np_AmQqUHHYC9dxwXRRYgt30EwA,2128
-torchrir/signal.py,sha256=QETI2dv5CPU4H9Jpi4CSEow4qlYVn0kbsun02h9MnP8,8253
-torchrir/simulators.py,sha256=4hBswrBc8erp4ac5O20YbcVN8V_OnjP5ANaVnOtJQ9E,3075
-torchrir/utils.py,sha256=HNQDksskfk6xLZ7XFWA7mh9GHhdfW4SthnJ5w_k58_Y,10652
-torchrir/datasets/__init__.py,sha256=3T55F3fjjRR3j618ubRkMlZnQTxvXaxioFMhygxm7oQ,601
-torchrir/datasets/base.py,sha256=mCHLtGOOaD1II1alJpP6ipzkz87l-rh19NgfeLnJbDU,720
-torchrir/datasets/cmu_arctic.py,sha256=NpF5fATptkBPjf-6fDUYrOqfl-Jvi3j9bj5wfvwmcLI,7034
-torchrir/datasets/template.py,sha256=KByYkCRm3cdTj4_jDvcZQ0Z4igilojJBH-6_W0-JIyc,2076
-torchrir/datasets/utils.py,sha256=UtG8fcBaHiJiZ8Pba9IBY5_ClA_K_ilqUpGzvnwGNpQ,3135
-torchrir-0.1.2.dist-info/licenses/LICENSE,sha256=5vS_7WTsMEw_QQHEPQ_WCwovJXEgmxoEwcwOI-9VbXI,10766
-torchrir-0.1.2.dist-info/licenses/NOTICE,sha256=SRs_q-ZqoVF9_YuuedZOvVBk01jV7YQAeF8rRvlRg0s,118
-torchrir-0.1.2.dist-info/METADATA,sha256=EWDdQ4sWDAcOgCDlg7b4QHFLu_Wm-6CXVc3sY-gx1E0,9161
-torchrir-0.1.2.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-torchrir-0.1.2.dist-info/top_level.txt,sha256=aIFwntowJjvm7rZk480HymC3ipDo1g-9hEbNY1wF-Oo,9
-torchrir-0.1.2.dist-info/RECORD,,