Repository navigation
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/25153
Note: Links to docs will display an error until the docs builds have been completed. ❌ 3 New FailuresAs of commit a7c1599 with merge base 351e938 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
|
@claude review effort=xhigh |
|
Claude finished @rascani's task in 4m 3s —— View job I found no correctness blockers, but every unsupported linear (dynamic batch dimension, runtime bias, pruned channels) now fails the whole export with no fallback, and the C++ kernel doesn't check the shape and dtype relationships between its arguments. Detailed reviewPR Review: #25153SummaryThis PR lowers
The Code Quality
Infrastructure
Testing
RecommendationNeeds Discussion. The main open question is the always-fail policy for rejected linears. Adding the kernel shape and dtype checks, and either allowing a dynamic batch dimension or documenting the restriction, would address the rest. |
4c8e856 to
857a90b
Compare
|
@claude /pr-review Please review the current revision ( |
|
Claude finished @rascani's task in 25m 32s —— View job At Detailed reviewPR Review: #25153SummaryThis PR lowers The supported path has no correctness blockers. The latest Test ESP32 Emulators run on this PR passed. That run covers ESP32-S3 with both the What remains:
Previous review findings, re-checked at this revision:
Code Quality
Testing
RecommendationNeeds Discussion. The supported path is numerically consistent across the pass, the reference op and the kernel, and passes the ESP32-S3 emulator tests. The open question is still the no-fallback policy for rejected linears. It is now documented, but it also leaves float I did not run any tests: Python and pytest need approval in this runner, so this review comes from reading the code. ESP-NN sources aren't vendored, so the ESP-NN side is verified only through the emulator CI result. |
857a90b to
3b908f8
Compare
fda524c to
0c4cb18
Compare
|
@claude /pr-review Please re-review the current revision ( This PR now enables the linear quantization rule together with the ESP-NN kernel; #25152 only prepares the rule. Please re-check the landing split and public recipe behavior. Unsupported FQ linear fallback remains an explicit existing limitation, while unquantized linear uses the portable decomposition. The focused quantizer and host linear checks passed. |
|
Claude finished @rascani's task in 29m 47s —— View job Re-review of Detailed reviewPR Review: #25153SummaryThis PR adds
Previous findings:
API Design
Code Quality
Testing
RecommendationNeeds Discussion: the open question is what the default recipe should do with linear layers it can tell before calibration that it cannot lower. It can raise, as now, or leave them float. Those models exported at base, and a
|
0c4cb18 to
b49ae09
Compare
cf22721 to
29b470c
Compare
Lower supported quantized linear calls to ESP-NN's batched per-channel fully connected API. Contiguous vector and higher-rank inputs use the row-based kernel without layout copies. Reuse the quantizer's int32 bias and prepare constant per-channel multipliers with bounds on integer accumulation and floating reference overflow. Unsupported calls use shared portable decomposition, including runtime weights or bias, oversized dimensions, partial quantization and unsafe accumulators. Strict export rejects those portable operators. The shared dialect and S3 cases cover accelerated and portable linear calls and serialize without remaining Fused Quant or aten.linear operators. Authored with OpenAI Codex.
Lower supported fused_quant.linear calls to ESP-NN's batched per-channel fully connected API. Static contiguous inputs of rank one or higher use the row-based kernel without layout copies, with constant symmetric per-channel int8 weights and optional derived int32 bias. LowerToEsp uses the stored bias and prepares bounded per-channel requantization parameters.
Unsupported calls use shared portable decomposition, including runtime weights or bias, oversized dimensions, partial quantization and unsafe accumulators. Strict export rejects the remaining portable compute. Serialized programs contain neither Fused Quant nor aten.linear operators.
Validation: all 28 host linear cases passed, including the focused portable-report rerun. Shared cases cover accelerated and portable calls, bias/no bias, vectors and higher ranks, aligned/tail widths, shared bias, float users, ReLU, chains, large bias and saturation. The fully composed stack also passed the runtime-bias and runtime-weight cases on the ESP32-S3 emulator. Scoped lint passed.
Authored with OpenAI Codex.