[ExecuTorch][WebGPU] Port clamp + hardtanh (clamp-family unary ops)#21156
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21156
Note: Links to docs will display an error until the docs builds have been completed. ❌ 5 New FailuresAs of commit c57c0c5 with merge base 4a26c64 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
This PR needs a
|
This was referenced Jul 22, 2026
This was referenced Jul 22, 2026
psiddh
approved these changes
Jul 23, 2026
psiddh
left a comment
psiddh
left a comment
Contributor
There was a problem hiding this comment.
Approving full WebGPU stack
meta-codesync
Bot
merged commit Jul 24, 2026
9feadfe
into
gh/JCNTH/116/base
182 of 191 checks passed
JCNTH
added a commit
that referenced
this pull request
Jul 24, 2026
Pull Request resolved: #21156 Ports the Vulkan delegate's clamp-family activations. `clamp` and `hardtanh` share one `clamp(x, min, max)` shader (mirroring how Vulkan reuses its clamp shader for both), built on the shared unary helper. Bounds come from the op's scalar args, with `None -> +/-inf` matching Vulkan `get_val_or_inf` (`backends/vulkan/runtime/graph/ops/impl/UnaryOp.cpp:72`). Key changes: - `runtime/ops/unary/clamp.wgsl` (+ generated `_wgsl.h`) — `output[idx] = clamp(input[idx], params.minimum, params.maximum)`; the `minimum`/`maximum` member names mirror Vulkan `unary_op.glsl`. - `runtime/ops/unary/Activations.cpp` — `clamp_impl` / `hardtanh_impl` (args `[in, min, max, out]`) + `get_val_or_inf` (reads an Int/Double scalar, substitutes `+/-inf` for `None`, throws on any other type). - `runtime/ops/unary/UnaryOp.cpp` — the shared helper now rejects int operands (fail-loud): the fp32-only backend would otherwise reinterpret int bits as f32, and `clamp` is the one unary the Vulkan partitioner (`op_registry.py`) tags for integer tensors. `relu6` is excluded: the Vulkan partitioner does not tag it, so it cannot be delegated without a separate partitioner change. ghstack-source-id: 406360830 @exported-using-ghexport Differential Revision: [D112257672](https://our.internmc.facebook.com/intern/diff/D112257672/)
JCNTH
added a commit
that referenced
this pull request
Jul 24, 2026
Pull Request resolved: #21156 Ports the Vulkan delegate's clamp-family activations. `clamp` and `hardtanh` share one `clamp(x, min, max)` shader (mirroring how Vulkan reuses its clamp shader for both), built on the shared unary helper. Bounds come from the op's scalar args, with `None -> +/-inf` matching Vulkan `get_val_or_inf` (`backends/vulkan/runtime/graph/ops/impl/UnaryOp.cpp:72`). Key changes: - `runtime/ops/unary/clamp.wgsl` (+ generated `_wgsl.h`) — `output[idx] = clamp(input[idx], params.minimum, params.maximum)`; the `minimum`/`maximum` member names mirror Vulkan `unary_op.glsl`. - `runtime/ops/unary/Activations.cpp` — `clamp_impl` / `hardtanh_impl` (args `[in, min, max, out]`) + `get_val_or_inf` (reads an Int/Double scalar, substitutes `+/-inf` for `None`, throws on any other type). - `runtime/ops/unary/UnaryOp.cpp` — the shared helper now rejects int operands (fail-loud): the fp32-only backend would otherwise reinterpret int bits as f32, and `clamp` is the one unary the Vulkan partitioner (`op_registry.py`) tags for integer tensors. `relu6` is excluded: the Vulkan partitioner does not tag it, so it cannot be delegated without a separate partitioner change. ghstack-source-id: 406360830 @exported-using-ghexport Differential Revision: [D112257672](https://our.internmc.facebook.com/intern/diff/D112257672/)
JCNTH
added a commit
that referenced
this pull request
Jul 24, 2026
Pull Request resolved: #21156 Ports the Vulkan delegate's clamp-family activations. `clamp` and `hardtanh` share one `clamp(x, min, max)` shader (mirroring how Vulkan reuses its clamp shader for both), built on the shared unary helper. Bounds come from the op's scalar args, with `None -> +/-inf` matching Vulkan `get_val_or_inf` (`backends/vulkan/runtime/graph/ops/impl/UnaryOp.cpp:72`). Key changes: - `runtime/ops/unary/clamp.wgsl` (+ generated `_wgsl.h`) — `output[idx] = clamp(input[idx], params.minimum, params.maximum)`; the `minimum`/`maximum` member names mirror Vulkan `unary_op.glsl`. - `runtime/ops/unary/Activations.cpp` — `clamp_impl` / `hardtanh_impl` (args `[in, min, max, out]`) + `get_val_or_inf` (reads an Int/Double scalar, substitutes `+/-inf` for `None`, throws on any other type). - `runtime/ops/unary/UnaryOp.cpp` — the shared helper now rejects int operands (fail-loud): the fp32-only backend would otherwise reinterpret int bits as f32, and `clamp` is the one unary the Vulkan partitioner (`op_registry.py`) tags for integer tensors. `relu6` is excluded: the Vulkan partitioner does not tag it, so it cannot be delegated without a separate partitioner change. ghstack-source-id: 406360830 @exported-using-ghexport Differential Revision: [D112257672](https://our.internmc.facebook.com/intern/diff/D112257672/)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
Ports the Vulkan delegate's clamp-family activations.
clampandhardtanhshare oneclamp(x, min, max)shader (mirroring how Vulkan reuses its clamp shader for both), built on the shared unary helper. Bounds come from the op's scalar args, withNone -> +/-infmatching Vulkanget_val_or_inf(backends/vulkan/runtime/graph/ops/impl/UnaryOp.cpp:72).Key changes:
runtime/ops/unary/clamp.wgsl(+ generated_wgsl.h) —output[idx] = clamp(input[idx], params.minimum, params.maximum); theminimum/maximummember names mirror Vulkanunary_op.glsl.runtime/ops/unary/Activations.cpp—clamp_impl/hardtanh_impl(args[in, min, max, out]) +get_val_or_inf(reads an Int/Double scalar, substitutes+/-infforNone, throws on any other type).runtime/ops/unary/UnaryOp.cpp— the shared helper now rejects int operands (fail-loud): the fp32-only backend would otherwise reinterpret int bits as f32, andclampis the one unary the Vulkan partitioner (op_registry.py) tags for integer tensors.relu6is excluded: the Vulkan partitioner does not tag it, so it cannot be delegated without a separate partitioner change.@exported-using-ghexport
Differential Revision: D112257672
Differential Revision: D112257672