Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages

, 'i'); if (__m === '*' || __re.test(location.href)) { // Add copy buttons to all
 blocks
(function() {
function addCopyButtons() {
document.querySelectorAll('pre code').forEach(function(codeBlock) {
if (codeBlock.parentElement.hasAttribute('data-copy-added')) return;
codeBlock.parentElement.setAttribute('data-copy-added', 'true');
var btn = document.createElement('button');
btn.textContent = 'Copy';
btn.style.cssText = 'position:absolute;top:4px;right:4px;padding:2px 8px;font-size:11px;background:#4ecdc4;border:none;border-radius:4px;color:#1a1a2e;cursor:pointer;opacity:0.7;transition:opacity 0.2s;';
btn.onmouseover = function() { this.style.opacity = '1'; };
btn.onmouseout = function() { this.style.opacity = '0.7'; };
btn.onclick = function() {
navigator.clipboard.writeText(codeBlock.textContent).then(function() {
btn.textContent = 'Copied!';
setTimeout(function() { btn.textContent = 'Copy'; }, 1500);
});
};
codeBlock.parentElement.style.position = 'relative';
codeBlock.parentElement.appendChild(btn);
});
}
addCopyButtons();
// Re-run on dynamic content
var observer = new MutationObserver(addCopyButtons);
observer.observe(document.body, { childList: true, subtree: true });
})();
}
} catch(__e) { console.warn('[Userscript:Add Copy Buttons to Code Blocks]', __e); }
})();
(function(){
try {
var __m = "github.com";
var __re = new RegExp('^' + "github\\.com" + '
GitHub - GOATnote-Inc/robogoat: GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo. · GitHub
Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages

, 'i'); if (__m === '*' || __re.test(location.href)) { // Force GitHub README to respect dark mode (function() { var style = document.createElement('style'); style.textContent = ' .markdown-body { color-scheme: dark light; } .markdown-body pre { background: #161b22 !important; } .markdown-body code { background: rgba(110, 118, 129, 0.4) !important; } .markdown-body table th, .markdown-body table td { border-color: #30363d !important; } .markdown-body img { background: #0d1117; } .markdown-body blockquote { border-left-color: #8b949e; } .markdown-body hr { border-color: #30363d; } '; document.head.appendChild(style); })(); } } catch(__e) { console.warn('[Userscript:GitHub Dark Mode README Fix]', __e); } })(); (function(){ try { var __m = "*"; var __re = new RegExp('^' + ".*" + ' GitHub - GOATnote-Inc/robogoat: GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo. · GitHub
Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages

, 'i'); if (__m === '*' || __re.test(location.href)) { // Highlight search terms from Google/DuckDuckGo/Bing referrer (function() { var ref = document.referrer; var terms = []; if (ref.includes('google.com') || ref.includes('duckduckgo.com') || ref.includes('bing.com')) { var url = new URL(ref); var q = url.searchParams.get('q') || url.searchParams.get('p'); if (q) { terms = q.split(/\s+/).filter(function(t) { return t.length > 2; }); } } if (terms.length === 0) return; var style = document.createElement('style'); style.textContent = '.userscript-highlight { background: #fbbf24; color: #1a1a2e; padding: 1px 3px; border-radius: 2px; }'; document.head.appendChild(style); function highlight(node) { if (node.nodeType === 3) { // text node var text = node.textContent; var found = false; terms.forEach(function(term) { var regex = new RegExp('(' + term.replace(/[.*+?^${}()|[\]\\]/g, '\\') + ')', 'gi'); if (regex.test(text)) { found = true; var frag = document.createDocumentFragment(); var parts = text.split(regex); parts.forEach(function(part, i) { if (i % 2 === 0) { frag.appendChild(document.createTextNode(part)); } else { var span = document.createElement('span'); span.className = 'userscript-highlight'; span.textContent = part; frag.appendChild(span); } }); node.parentNode.replaceChild(frag, node); } }); } else if (node.nodeType === 1 && node.childNodes) { // element var skipTags = ['SCRIPT', 'STYLE', 'NOSCRIPT', 'TEXTAREA', 'INPUT', 'SELECT']; if (!skipTags.includes(node.tagName)) { Array.from(node.childNodes).forEach(highlight); } } } highlight(document.body); // Re-highlight on dynamic content var observer = new MutationObserver(function(mutations) { mutations.forEach(function(m) { m.addedNodes.forEach(function(node) { if (node.nodeType === 1 || node.nodeType === 3) highlight(node); }); }); }); observer.observe(document.body, { childList: true, subtree: true }); })(); } } catch(__e) { console.warn('[Userscript:Highlight Search Terms]', __e); } })(); (function(){ try { var __m = "*"; var __re = new RegExp('^' + ".*" + ' GitHub - GOATnote-Inc/robogoat: GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo. · GitHub
Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages

, 'i'); if (__m === '*' || __re.test(location.href)) { // Strip utm_, fbclid, gclid, etc. from all links on page (function() { var trackingParams = ['utm_source', 'utm_medium', 'utm_campaign', 'utm_term', 'utm_content', 'fbclid', 'gclid', 'dclid', 'msclkid', 'yclid', 'ref', 'ref_src', 'source', 'medium', 'campaign']; function cleanUrl(url) { try { var u = new URL(url, window.location.origin); var changed = false; trackingParams.forEach(function(p) { if (u.searchParams.has(p)) { u.searchParams.delete(p); changed = true; } }); return changed ? u.toString() : url; } catch (e) { return url; } } function cleanLinks() { document.querySelectorAll('a[href]').forEach(function(a) { var clean = cleanUrl(a.href); if (clean !== a.href) a.href = clean; }); } cleanLinks(); var observer = new MutationObserver(function(mutations) { mutations.forEach(function(m) { m.addedNodes.forEach(function(node) { if (node.nodeType === 1) { if (node.tagName === 'A') cleanLinks(); node.querySelectorAll('a[href]').forEach(function(a) { var clean = cleanUrl(a.href); if (clean !== a.href) a.href = clean; }); } }); }); }); observer.observe(document.body, { childList: true, subtree: true }); })(); } } catch(__e) { console.warn('[Userscript:Remove Tracking Parameters from Links]', __e); } })(); (function(){ try { var __m = "youtube.com"; var __re = new RegExp('^' + "youtube\\.com" + ' GitHub - GOATnote-Inc/robogoat: GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo. · GitHub
Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages

, 'i'); if (__m === '*' || __re.test(location.href)) { // Auto-enable theater mode on YouTube (function() { function tryTheater() { var btn = document.querySelector('button[aria-label="Theater mode"], ytd-player #player button[title="Theater mode"]'); if (btn && !btn.classList.contains('activated')) { btn.click(); } } // Try immediately tryTheater(); // Try after navigation (SPA) var lastUrl = location.href; setInterval(function() { if (location.href !== lastUrl) { lastUrl = location.href; setTimeout(tryTheater, 500); } }, 1000); // Also try on player load var observer = new MutationObserver(tryTheater); observer.observe(document.body, { childList: true, subtree: true }); })(); } } catch(__e) { console.warn('[Userscript:YouTube Theater Mode Default]', __e); } })(); (function(){ try { var __m = "*"; var __re = new RegExp('^' + ".*" + ' GitHub - GOATnote-Inc/robogoat: GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo. · GitHub
Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages

, 'i'); if (__m === '*' || __re.test(location.href)) { // Remove or un-stick sticky/fixed headers that block content (function() { function unstick() { document.querySelectorAll('header, nav, [role="banner"], .header, .navbar, .sticky, .fixed-top, [style*="position: fixed"], [style*="position:sticky"]').forEach(function(el) { if (el.style.position === 'fixed' || el.style.position === 'sticky' || getComputedStyle(el).position === 'fixed' || getComputedStyle(el).position === 'sticky') { el.style.position = 'static'; el.style.top = 'auto'; el.style.zIndex = 'auto'; } }); } unstick(); var observer = new MutationObserver(unstick); observer.observe(document.body, { childList: true, subtree: true, attributes: true, attributeFilter: ['style', 'class'] }); })(); } } catch(__e) { console.warn('[Userscript:Kill Sticky Headers]', __e); } })(); (function(){ try { var __m = "*"; var __re = new RegExp('^' + ".*" + ' GitHub - GOATnote-Inc/robogoat: GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo. · GitHub
Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages

, 'i'); if (__m === '*' || __re.test(location.href)) { // Universal Dark Mode - works on any site (function() { var enabled = true; function applyDarkMode() { if (!enabled) return; // Create style element if it doesn't exist var style = document.getElementById('universal-dark-mode-style'); if (!style) { style = document.createElement('style'); style.id = 'universal-dark-mode-style'; document.head.appendChild(style); } // Dark mode CSS - inverts colors but preserves images/video style.textContent = ' /* Invert everything except media */ html { filter: invert(1) hue-rotate(180deg) !important; background: #1a1a2e !important; } /* Restore images, videos, iframes, canvas */ img, video, iframe, canvas, svg, picture, [style*="background-image"] { filter: invert(1) hue-rotate(180deg) !important; } /* Preserve specific elements that should not be inverted */ .no-dark-mode, .no-dark-mode *, [data-theme="light"], [data-theme="light"], .ace_editor, .ace_editor *, .CodeMirror, .CodeMirror *, .monaco-editor, .monaco-editor *, .markdown-body pre, .markdown-body pre *, .highlight, .highlight *, pre code, pre code * { filter: none !important; } /* Fix common UI elements */ .modal, .popup, .dropdown-menu, .tooltip, .popover { filter: invert(1) hue-rotate(180deg) !important; background: #2d2d44 !important; border-color: #444 !important; } /* Scrollbars */ ::-webkit-scrollbar { background: #1a1a2e !important; } ::-webkit-scrollbar-thumb { background: #444 !important; } ::-webkit-scrollbar-thumb:hover { background: #555 !important; } /* Selection */ ::selection { background: #4ecdc4 !important; color: #1a1a2e !important; } ::-moz-selection { background: #4ecdc4 !important; color: #1a1a2e !important; } '; } function removeDarkMode() { var style = document.getElementById('universal-dark-mode-style'); if (style) style.remove(); } // Toggle with Alt+Shift+D document.addEventListener('keydown', function(e) { if (e.altKey && e.shiftKey && e.key === 'D') { e.preventDefault(); enabled = !enabled; if (enabled) { applyDarkMode(); console.log('[Universal Dark Mode] Enabled'); } else { removeDarkMode(); console.log('[Universal Dark Mode] Disabled'); } } }); // Apply on load applyDarkMode(); // Re-apply on dynamic content var observer = new MutationObserver(function(mutations) { if (enabled && !document.getElementById('universal-dark-mode-style')) { applyDarkMode(); } }); observer.observe(document.head, { childList: true }); console.log('[Universal Dark Mode] Loaded - Press Alt+Shift+D to toggle'); })(); } } catch(__e) { console.warn('[Userscript:Universal Dark Mode]', __e); } })(); })(); GitHub - GOATnote-Inc/robogoat: GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo. · GitHub
Skip to content
This repository was archived by the owner on Aug 31, 2026. It is now read-only.

Latest commit

History

218 Commits

Folders and files

NameName
Last commit message
Last commit date

RoboCache

Archived August 2026. Last GPU validation 2025-11-08 on H100 PCIe at commit 0db3726; numbers below are from that run and have not been re-validated.

GPU-Accelerated Data Engine for Robot Foundation Models

LicenseCUDAPythonPyTorch

Quick Start | Installation | Performance | Documentation


Overview

RoboCache is a CUDA library for sensor preprocessing in robotics: GPU-accelerated temporal alignment (trajectory resampling, multimodal fusion) and point cloud voxelization, with a pure-PyTorch fallback.

Scope of validation:

  • CUDA kernels were benchmarked on H100 PCIe and A100 SXM4 in November 2025.
  • All benchmark inputs were synthetic tensors (torch.randn / torch.rand), including the "dataset" benchmarks, which use tensors shaped like Isaac Gym, TartanAir, nuScenes, and KITTI samples - no real dataset files were loaded.
  • The CPU fallback has known correctness bugs (see Known Issues).

Quick Start

importtorchimportrobocache# 3-stream multimodal fusion (vision + proprioception + IMU)vision=torch.randn(4, 30, 512, dtype=torch.bfloat16, device='cuda')
vision_times=torch.linspace(0, 1, 30, device='cuda').expand(4, -1)
proprio=torch.randn(4, 100, 64, dtype=torch.bfloat16, device='cuda')
proprio_times=torch.linspace(0, 1, 100, device='cuda').expand(4, -1)
imu=torch.randn(4, 200, 12, dtype=torch.bfloat16, device='cuda')
imu_times=torch.linspace(0, 1, 200, device='cuda').expand(4, -1)
target_times=torch.linspace(0, 1, 50, device='cuda').expand(4, -1)
# Fuse all streams to common timelinefused=robocache.fuse_multimodal(
vision, vision_times,
proprio, proprio_times,
imu, imu_times,
target_times
)
# Output: (4, 50, 588) - batch x time x (512+64+12)

Point Cloud Voxelization:

# LiDAR -> 3D voxel gridpoints=torch.rand(500000, 3, device='cuda') *20.0-10.0voxel_grid=robocache.voxelize_pointcloud(
points,
grid_min=[-10.0, -10.0, -10.0],
voxel_size=0.05, # 5cm voxelsgrid_size=[128, 128, 128],
mode='occupancy'
)

Installation

Not published to PyPI. Install from source:

git clone https://github.com/GOATnote-Inc/robogoat.git
cd robogoat/robocache
# Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu121
# Build CUDA extensions
python setup.py develop
# Verify
python -c "import robocache; robocache.self_test()"

Requirements:

  • NVIDIA GPU (Compute Capability >= 8.0)
  • CUDA 12.1+ or 13.0+
  • PyTorch 2.0+

Performance

All numbers are from the November 2025 validation run on a single NVIDIA H100 PCIe 80GB (CUDA 13.0, driver 580.95) and have not been re-validated since.

Kernel microbenchmarks (H100, CUDA kernel vs. PyTorch on CPU)

Source: robocache/bench/results/benchmark_h100_20251106_172811.csv (5 seeds x 50 repeats = 250 measurements per CUDA config; synthetic bf16 tensors).

Trajectory resample config (B x S -> T, D)CUDA P50PyTorch CPU P50Speedup
8 x 250, 128 (small)0.184 ms20.14 ms~110x
32 x 500, 256 (medium)2.605 ms38.39 ms~15x
64 x 1000, 512 (large)20.05 ms75.69 ms~3.8x

These are kernel-vs-CPU microbenchmarks, not end-to-end training comparisons.

End-to-end training (H100, measured)

Source: PRODUCTION_STATUS.md and robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md.

Pipelinems/stepSpeedup
Baseline (PyTorch preprocessing)18.281.00x
RoboCache preprocessing14.041.30x

The measured end-to-end training speedup is 1.30x, driven by preprocessing being a minority of step time once the model forward/backward is included.

Known regression (measured, documented)

Source: robocache/benchmarks/results/h100_validated_20251105.json.

Config (B x S -> T, D)RoboCachePyTorch GPUResult
64 x 4096 -> 1024, 320.190 ms0.140 ms0.74x (slower)

For long sequences (> ~2000 timesteps) the per-thread binary search falls out of L1 cache and native PyTorch GPU interpolation is faster. See KNOWN_LIMITATIONS.md.

Profiling artifacts

Nsight Compute and Nsight Systems text captures from the H100/A100 runs are committed under artifacts/h100/ and artifacts/a100/ (with GPU and driver stamps). Binary .ncu-rep / .nsys-rep files are not committed.


Examples


Testing

cd robocache
pytest tests/ -v

CI: Lint + CPU tests on every PR (.github/workflows/ci.yml). CUDA kernels are not exercised in CI; GPU validation was manual (see docs/validation/). The self-hosted GPU runners used for that validation no longer exist.


Known Issues

  • CPU voxelization fallback is incorrect.ops_fallback.voxelize_pointcloud_cpu clamps out-of-bounds points into boundary voxels instead of dropping them (diverging from the CUDA kernel), and raises IndexError on empty or single-point clouds. The corresponding CPU tests are marked xfail. Do not use the CPU fallback where voxel occupancy correctness matters.
  • Stale multimodal-fusion tests.robocache/tests/test_multimodal_fusion.py targets a pre-1.0 two-stream fuse_multimodal API and is skipped; the current API takes three streams.
  • Compute Sanitizer / 24h burn-in were never run in CI. The stress-test code exists but there is no committed memcheck/racecheck log.
  • Voxelization out-of-bounds behavior (CUDA): points outside the grid are clipped, no error is thrown.
  • Timestamp monotonicity is not enforced; callers must supply monotonically increasing timestamps.

Documentation


Citation

@software{robocache2025,
title={RoboCache: GPU-Accelerated Data Engine for Robot Learning},
author={Dent, Brandon},
year={2025},
url={https://github.com/GOATnote-Inc/robogoat}
}

License

Apache 2.0 - See LICENSE


Maintained by:GOATnoteStatus: Archived (August 2026)

About

GPU data-loading/preprocessing kernels for robot learning research. Archived 2026-08; last GPU validation 2025-11-08 (H100 PCIe). Measured speedups documented in-repo.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages