Repository navigation
Conversation
Accelerate static int8 means over the spatial dimensions of 4-D tensors with independent input and output quantization parameters. Convert the input to NHWC, run the vendor mean kernel per batch, and restore singleton dimensions for keepdim. Existing permutation cleanup removes redundant conversions after spatial producers. Unsupported reductions and unsafe arithmetic retain portable execution. Cover lowering, serialization, layout cleanup, arithmetic bounds, strict export, and shared ESP32-S3 cases. Refresh the MLPerf Tiny expectations to accelerate the means in ResNet-8 and MobileNetV1 0.25. Validation: 176 Python checks, four model coverage/serialization tests, 19 ESP32-S3 emulator cases, and changed-file lint passed. Authored with OpenAI Codex.
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/25405
Note: Links to docs will display an error until the docs builds have been completed. ❌ 3 Cancelled Jobs, 1 Pending, 1 Unrelated FailureAs of commit 735aedb with merge base 351e938 ( CANCELLED JOBS - The following jobs were cancelled. Please retry:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
Accelerate static int8 means over the spatial dimensions of 4-D tensors with independent input and output quantization parameters. Convert the input to NHWC, run the vendor mean kernel per batch, and restore singleton dimensions for keepdim. Existing permutation cleanup removes redundant conversions after spatial producers. Unsupported reductions and unsafe arithmetic retain portable execution.
Cover lowering, serialization, layout cleanup, arithmetic bounds, strict export, and shared ESP32-S3 cases. Refresh the MLPerf Tiny expectations to accelerate the means in ResNet-8 and MobileNetV1 0.25. Validation: 176 Python checks, four model coverage/serialization tests, 19 ESP32-S3 emulator cases, and changed-file lint passed.
Authored with OpenAI Codex.