[claude] Make MXFP8 cuda kernels ABI stable by andrewor14 · Pull Request #3610 · pytorch/ao

andrewor14 · 2026-01-09T16:25:57Z

Test Plan:

pytest test/prototype/mx_formats/test_kernels.py -v -s -k cuda
pytest test/prototype/moe_training/test_kernels.py -v -s -k cuda_mx_block
pytest test/prototype/moe_training/test_training.py -v -s

Prompt:

Make these two files ABI stable:
torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu
torchao/csrc/cuda/mx_kernels/mxfp8_extension.cpp

Read these for instructions:
pytorch/docs/source/notes/libtorch_stable_abi.md
cppdocs/_sources/stable.rst.txt

Use these files for an example:
Before: flash-attention/hopper/flash_api.cpp
After: flash-attention/hopper/flash_api_stable.cpp

Additional instructions:
Replace at::zeros with torch::stable::new_zeros
Replace at::empty with torch::stable::new_empty
Replace TORCH_CHECK with STD_TORCH_CHECK without changing the content of the check, just replace the function name
Replace c10::cuda::CUDAGuard with DeviceGuard
When calling tensor.stride(), no need to check for tensor.numel() > 0 beforehand
When using cudaStream_t, remember to include cuda_runtime.h
When calling aoti_torch_get_current_cuda_stream, get the device index from a tensor, not from the general torch::stable::accelerator::getCurrentDeviceIndex();

Don’t do these things:
Don’t use torch::stable::new_empty_strided since this API does not exist
Don’t define a dummy _C module that can be accessed from python
Don’t declare aoti_torch_get_current_cuda_stream, just include it from torch/csrc/inductor/aoti_torch/c/shim.h and add -DUSE_CUDA to both cxx and nvcc in setup.py
Don’t box kernels manually, just use TORCH_BOX
Don’t change switch statements into if else blocks

Follow-up prompts:

Revert the content changes of STD_TORCH_CHECK, just replace the function name

For the changes that involved tensor.data_ptr<some_dtype>() in the old code, revert the changes and replace it with tensor.const_data_ptr<some_dtype>()

Output:

● I've successfully made both files ABI stable. Here's a summary of the changes:

  Changes to mxfp8_cuda.cu:

  1. Updated includes: Replaced ATen and torch/extension.h with stable ABI headers (torch/csrc/stable/tensor.h, torch/csrc/inductor/aoti_torch/c/shim.h, torch/headeronly/core/ScalarType.h, torch/headeronly/util/Exception.h, torch/headeronly/util/shim_utils.h)
  2. Added cuda_runtime.h for cudaStream_t
  3. Changed tensor type: torch::Tensor → torch::stable::Tensor (with using declaration)
  4. Changed scalar type enums: torch::kFloat32 → torch::headeronly::ScalarType::Float, etc.
  5. Changed error checking: TORCH_CHECK → STD_TORCH_CHECK
  6. Changed CUDA stream access: at::cuda::getCurrentCUDAStream() → aoti_torch_get_current_cuda_stream(input.get_device_index(), &stream_ptr) with TORCH_ERROR_CODE_CHECK
  7. Fixed stride access: .strides()[i] → .stride(i) and .sizes()[i] → .size(i)

  Changes to mxfp8_extension.cpp:

  1. Updated includes: Replaced torch/library.h, ATen/ATen.h, ATen/cuda/CUDAContext.h, c10/cuda/CUDAGuard.h with stable headers
  2. Changed tensor type: at::Tensor → torch::stable::Tensor
  3. Changed DeviceGuard: c10::cuda::CUDAGuard → torch::stable::accelerator::DeviceGuard (with input.get_device_index() instead of input.device())
  4. Changed error checking: TORCH_CHECK → STD_TORCH_CHECK
  5. Changed tensor creation:
    - at::empty → torch::stable::new_empty
    - at::zeros → torch::stable::new_zeros
    - For strided tensors: Created with transposed dimensions then used torch::stable::transpose to get column-major layout
  6. Changed CUDA stream access: Used aoti_torch_get_current_cuda_stream with device index from tensor
  7. Changed library registration: TORCH_LIBRARY_IMPL → STABLE_TORCH_LIBRARY_IMPL with TORCH_BOX wrapper

  Changes to setup.py:

  1. Added -DUSE_CUDA to both cxx and nvcc compiler arguments for the mxfp8 extension

pytorch-bot · 2026-01-09T16:26:01Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3610

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⏳ No Failures, 18 Pending

As of commit 0211669 with merge base 34322b5 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

janeyx99

The best way to test this is by setting up a build where you can pass in -DTORCH_TARGET_VERSION and build the kernel + ensure it builds and runs. I'd recommend setting up a separate stable wheel for the migration to make testing easier as you slowly migrate kernels to the stable wheel

Edit: I just learned that you only have 3 files to migrate. In that case you do have the option of just biting the bullet and changing the setup.py in torchao to add this flag: https://github.com/Dao-AILab/flash-attention/pull/2155/files.

Note you want 20a0 and not 2090 though, for 2.10.

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu

janeyx99 · 2026-01-09T20:43:54Z

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu

+#include <torch/headeronly/util/Exception.h>
+
+// Declare the CUDA stream function that's behind #ifdef USE_CUDA in shim.h
+extern "C" AOTITorchError aoti_torch_get_current_cuda_stream(int32_t device_index, void** ret_stream);


I don't think you need this, you should just pass -DUSE_CUDA as a flag in setup.py.

Though hmm @mikaylagawarecki should we automatically pass this in through CUDAExtension for better UX?

This or -DUSE_CUDA would both work

tbh I'm not sure that silently defining USE_CUDA in CUDAExtension is good UX 🤔 (Where did USE_CUDA originate from? Is it a pytorch thing or a cuda thing? I'm wondering whether users might also define USE_CUDA themselves for separate purposes hmm, wdyt?)

janeyx99 · 2026-01-09T20:44:38Z

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu

-  case torch::kFloat32:
+DType get_input_dtype(const Tensor &t) {
+  auto scalar_type = t.scalar_type();
+  if (scalar_type == torch::headeronly::ScalarType::Float) {


you should still be able to use a switch statement i believe, but no biggie

janeyx99 · 2026-01-09T20:46:31Z

torchao/csrc/cuda/mx_kernels/mxfp8_extension.cpp

    const int64_t num_col_blocks = (cols + scale_dim_x - 1) / scale_dim_x;
-    output_rowwise = at::empty({rows, cols}, options_fp8);
-    scales_rowwise = at::empty({rows, num_col_blocks}, options_scale);
+    output_rowwise = torch::stable::empty(


new_empty?

here and below

janeyx99

Yea, this looks close. Here's when I'd recommend just building the wheel and letting the error messages and test cases get you 100% there.

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu

andrewor14 · 2026-01-15T22:14:41Z

torchao/csrc/cuda/mx_kernels/mxfp8_extension.cpp

+    // Create column-major tensor by creating transposed shape and transposing
+    // We need shape {rows, cols} with strides {1, rows}, so create {cols, rows} and transpose
+    Tensor output_colwise_tmp = torch::stable::new_empty(input, {cols, rows}, torch::headeronly::ScalarType::Float8_e4m3fn);
+    output_colwise = torch::stable::transpose(output_colwise_tmp, 0, 1);


@janeyx99 @danielvegamyhre is this right? Seems like claude is doing this for all column major tensors

looks right to me, it allocates empty shape (cols, rows) with strides (rows, 1) (row major) then transpose so we have (rows, cols) with strides (1, rows) (col major). since transpose is just a metadata change this shouldn't impact perf

also as an aside (and a note to myself) i think the original code for the scales can be simplified, but will leave it for a future PR to avoid merge conflicts with this

andrewor14 · 2026-01-15T22:15:49Z

torchao/csrc/cuda/mx_kernels/mxfp8_extension.cpp

  // Get raw pointers
  const uint8_t* scales_ptr = reinterpret_cast<const uint8_t*>(scales_tensor.data_ptr());
-  const int32_t* offsets_ptr = input_group_end_offsets.data_ptr<int32_t>();
+  const int32_t* offsets_ptr = static_cast<const int32_t*>(input_group_end_offsets.data_ptr());


do we need this change?

You can use input_group_end_offsets.const_data_ptr<int32_t>() the stable Tensor.data_ptr() doesn't accept templates atm hence the issue :(

for future reference, there's also .mutable_data_ptr() in stable

andrewor14 · 2026-01-15T22:16:18Z

setup.py

                        + [
                            "-gencode=arch=compute_100,code=sm_100",
                            "-gencode=arch=compute_120,code=compute_120",
+                            "-DUSE_CUDA",


@janeyx99 @mikaylagawarecki I needed to add this for aoti_torch_get_current_cuda_stream. Is this what you would recommend?

janeyx99

Approval contingent on passing in the build flag and the newly built AO wheel being able to run tests correctly.

janeyx99 · 2026-01-26T20:16:14Z

setup.py

                            f"-DPy_LIMITED_API={min_supported_cpython_hexcode}",
                            "-std=c++17",
                            "-O3",
+                            "-DUSE_CUDA",


also pass in a -DTORCH_TARGET_VERSION=0x020a000000000000 for 2.10

danielvegamyhre · 2026-01-26T23:11:25Z

@andrewor14 i'll build and do some tests with this PR locally and follow up with the results

danielvegamyhre · 2026-01-26T23:43:14Z

@andrewor14 when I run unit tests I see runtime errors like this:

RuntimeError: call, /home/dev/.conda/envs/torch/lib/python3.12/site-packages/torch/include/torch/csrc/stable/stableivalue_conversions.h:530, Not yet supported ScalarType 44, please file an issue describing your use case.

Repro command:

pytest test/prototype/moe_training/test_training.py -v -s

@janeyx99 any thoughts on this?

danielvegamyhre · 2026-01-27T00:33:15Z

also @andrewor14 i just landed a test fix for an unrelated issue #3728, i would suggest rebasing on main now to get cleaner test signal

andrewor14 · 2026-01-27T15:51:01Z

Hmm I'm hitting a different error after rebasing and running the test:

pytest test/prototype/moe_training/test_training.py -v -s

Error:

FAILED test/prototype/moe_training/test_training.py::test_moe_training[recipe_config1-False-target_fqns0] -

RuntimeError: CUDA error: CUBLAS_STATUS_INVALID_VALUE when calling
`cublasGemmEx( handle, opa, opb, m, n, k, &falpha, a, CUDA_R_16BF, lda,
b, CUDA_R_16BF, ldb, &fbeta, c, std::is_same_v<C_Dtype, float...

Any ideas? Wonder if it's a shape error? We did add some transposes in this PR. @danielvegamyhre

janeyx99 · 2026-01-27T16:09:08Z

RuntimeError: call, /home/dev/.conda/envs/torch/lib/python3.12/site-packages/torch/include/torch/csrc/stable/stableivalue_conversions.h:530, Not yet supported ScalarType 44, please file an issue describing your use case.
Repro command:

pytest test/prototype/moe_training/test_training.py -v -s

@janeyx99 any thoughts on this?

Oh boy, this is cuz our shim doesn't currently support the 44th ScalarType that is Float8_e8m0fnu --> is that needed? I'll patch something for 2.11 if so, but it would mean we can't migrate this til then :/

andrewor14 · 2026-01-28T16:54:45Z

Marking this PR as draft so we don't accidentally merge it. Seems like this is blocked on Float8_e8m0fnu ABI stable support for now, probably until torch 2.11

andrewor14 · 2026-02-06T20:13:35Z

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu

+  int64_t scale_rowwise_stride_dim0 = scales_rowwise.dim() >= 2 ? scales_rowwise.stride(0) : 0;
+  int64_t scale_rowwise_stride_dim1 = scales_rowwise.dim() >= 2 ? scales_rowwise.stride(1) : 0;
+  int64_t scale_colwise_stride_dim0 = scales_colwise.dim() >= 2 ? scales_colwise.stride(0) : 0;
+  int64_t scale_colwise_stride_dim1 = scales_colwise.dim() >= 2 ? scales_colwise.stride(1) : 0;


Hey @danielvegamyhre I had to add the dim check to pass tests for some reason. Does this look right? Here's what claude said:

The Problem: In your commit, you changed from .strides()[i] to .stride(i) to use the stable ABI. However, the stable ABI's .stride(i) method validates the dimension index, while .strides()[i] accessed the array directly without bounds checking. When rowwise=False, scales_rowwise is created as a 1D tensor with shape {0}. When the code calls scales_rowwise.stride(1), it fails because dimension index 1 is out of range for a 1D tensor (valid indices are only -1 and 0). The Fix (in torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu): Guard the stride accesses with dimension checks: // Get strides of scale ptrs (guard against 1D empty tensors when rowwise/colwise is false) int64_t scale_rowwise_stride_dim0 = scales_rowwise.dim() >= 2 ? scales_rowwise.stride(0) : 0; int64_t scale_rowwise_stride_dim1 = scales_rowwise.dim() >= 2 ? scales_rowwise.stride(1) : 0; int64_t scale_colwise_stride_dim0 = scales_colwise.dim() >= 2 ? scales_colwise.stride(0) : 0; int64_t scale_colwise_stride_dim1 = scales_colwise.dim() >= 2 ? scales_colwise.stride(1) : 0; This was actually a latent bug in the original code - the unstable ABI's .strides()[i] silently returned garbage/undefined values for out-of-bounds access. The stable ABI's stricter bounds checking exposed it.

Looks good to me

andrewor14 · 2026-02-06T20:14:45Z

setup.py

                            "-O3",
+                            "-DUSE_CUDA",
+                            # define TORCH_TARGET_VERSION with min version 2.11 for ABI stable Float8_e8m0fnu
+                            "-DTORCH_TARGET_VERSION=0x020b000000000000",


Note: updated the target version to 2.11 cause we need pytorch/pytorch#173669

should we add a guard for torch_version_at_least("2.11.0") as well?

Prompt: ``` I want you to make the mxfp8 kernels under torchao/csrc/cuda/mx_kernels ABI stable. Read these for instructions: pytorch/docs/source/notes/libtorch_stable_abi.md cppdocs/_sources/stable.rst.txt Use these files for an example: Before: flash-attention/hopper/flash_api.cpp After: flash-attention/hopper/flash_api_stable.cpp Additional instructions: Replace at::zeros with torch::stable::new_zeros Replace at::empty with torch::stable::new_empty Replace TORCH_CHECK with STD_TORCH_CHECK without changing the content of the check, just replace the function name Replace c10::cuda::CUDAGuard with DeviceGuard When calling tensor.stride(), no need to check for tensor.numel() > 0 beforehand When using cudaStream_t, remember to include cuda_runtime.h When calling aoti_torch_get_current_cuda_stream, get the device index from a tensor, not from the general torch::stable::accelerator::getCurrentDeviceIndex(); Don’t do these things: Don’t use torch::stable::new_empty_strided since this API does not exist Don’t define a dummy _C module that can be accessed from python Don’t declare aoti_torch_get_current_cuda_stream, just include it from torch/csrc/inductor/aoti_torch/c/shim.h and add -DUSE_CUDA to both cxx and nvcc in setup.py Don’t box kernels manually, just use TORCH_BOX Don’t change switch statements into if else blocks ``` Follow-up prompts: ``` Revert the content changes of STD_TORCH_CHECK, just replace the function name For the changes that involved tensor.data_ptr<some_dtype>() in the old code, revert the changes and replace it with tensor.const_data_ptr<some_dtype>() ``` Output: ``` ● I've successfully made both files ABI stable. Here's a summary of the changes: Changes to mxfp8_cuda.cu: 1. Updated includes: Replaced ATen and torch/extension.h with stable ABI headers (torch/csrc/stable/tensor.h, torch/csrc/inductor/aoti_torch/c/shim.h, torch/headeronly/core/ScalarType.h, torch/headeronly/util/Exception.h, torch/headeronly/util/shim_utils.h) 2. Added cuda_runtime.h for cudaStream_t 3. Changed tensor type: torch::Tensor → torch::stable::Tensor (with using declaration) 4. Changed scalar type enums: torch::kFloat32 → torch::headeronly::ScalarType::Float, etc. 5. Changed error checking: TORCH_CHECK → STD_TORCH_CHECK 6. Changed CUDA stream access: at::cuda::getCurrentCUDAStream() → aoti_torch_get_current_cuda_stream(input.get_device_index(), &stream_ptr) with TORCH_ERROR_CODE_CHECK 7. Fixed stride access: .strides()[i] → .stride(i) and .sizes()[i] → .size(i) Changes to mxfp8_extension.cpp: 1. Updated includes: Replaced torch/library.h, ATen/ATen.h, ATen/cuda/CUDAContext.h, c10/cuda/CUDAGuard.h with stable headers 2. Changed tensor type: at::Tensor → torch::stable::Tensor 3. Changed DeviceGuard: c10::cuda::CUDAGuard → torch::stable::accelerator::DeviceGuard (with input.get_device_index() instead of input.device()) 4. Changed error checking: TORCH_CHECK → STD_TORCH_CHECK 5. Changed tensor creation: - at::empty → torch::stable::new_empty - at::zeros → torch::stable::new_zeros - For strided tensors: Created with transposed dimensions then used torch::stable::transpose to get column-major layout 6. Changed CUDA stream access: Used aoti_torch_get_current_cuda_stream with device index from tensor 7. Changed library registration: TORCH_LIBRARY_IMPL → STABLE_TORCH_LIBRARY_IMPL with TORCH_BOX wrapper Changes to setup.py: 1. Added -DUSE_CUDA to both cxx and nvcc compiler arguments for the mxfp8 extension ```

andrewor14 requested a review from danielvegamyhre January 9, 2026 16:26

meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jan 9, 2026

andrewor14 requested a review from janeyx99 January 9, 2026 16:26

andrewor14 changed the title ~~[draft] Make MXFP8 cuda kernel ABI stable~~ [claude] Make MXFP8 cuda kernel ABI stable Jan 9, 2026

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from 2f3e96d to 7b28a33 Compare January 9, 2026 16:26

andrewor14 added the topic: improvement Use this tag if this PR is an improvement (doesn't fit into any of the other categories) label Jan 9, 2026

janeyx99 reviewed Jan 9, 2026

View reviewed changes

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu Outdated Show resolved Hide resolved

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu Outdated Show resolved Hide resolved

andrewor14 marked this pull request as draft January 9, 2026 16:31

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch 2 times, most recently from 3784e4f to 2b13608 Compare January 9, 2026 20:29

andrewor14 changed the title ~~[claude] Make MXFP8 cuda kernel ABI stable~~ [claude] Make MXFP8 cuda kernels ABI stable Jan 9, 2026

andrewor14 requested a review from janeyx99 January 9, 2026 20:36

janeyx99 reviewed Jan 9, 2026

View reviewed changes

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch 3 times, most recently from ca53f6c to 70a834b Compare January 12, 2026 23:45

andrewor14 commented Jan 12, 2026

View reviewed changes

torchao/csrc/cuda/mx_kernels/mxfp8_cuda.cu Outdated Show resolved Hide resolved

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from 70a834b to 476167e Compare January 15, 2026 22:13

andrewor14 requested review from janeyx99 and mikaylagawarecki January 15, 2026 22:13

andrewor14 commented Jan 15, 2026

View reviewed changes

andrewor14 marked this pull request as ready for review January 15, 2026 22:19

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from 476167e to ff1b866 Compare January 16, 2026 18:58

andrewor14 mentioned this pull request Jan 26, 2026

Making torchao ABI stable and moving closer to python only #3516

Closed

janeyx99 approved these changes Jan 26, 2026

View reviewed changes

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from ff1b866 to 4fa9a70 Compare January 26, 2026 20:58

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from 4fa9a70 to 0841c49 Compare January 27, 2026 15:28

andrewor14 marked this pull request as draft January 28, 2026 16:54

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from 0841c49 to 5edc119 Compare February 6, 2026 20:11

andrewor14 commented Feb 6, 2026

View reviewed changes

andrewor14 mentioned this pull request Feb 6, 2026

Support Float8_e8m0fnu and Float4_e2m1fn_x2 in stable ABI pytorch/pytorch#173669

Closed

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from 5edc119 to c19db71 Compare March 19, 2026 22:00

andrewor14 added the module: not user facing Use this tag if you don't want this PR to show up in release notes label Mar 19, 2026

andrewor14 marked this pull request as ready for review March 19, 2026 22:01

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch 2 times, most recently from 8558dff to a4fca0c Compare March 23, 2026 21:20

andrewor14 force-pushed the mxfp8-kernel-abi-stable branch from a4fca0c to 0211669 Compare March 23, 2026 21:21

andrewor14 merged commit aa9be68 into main Mar 23, 2026
27 of 33 checks passed

Conversation

andrewor14 commented Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3610

⏳ No Failures, 18 Pending

Uh oh!

janeyx99 left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mikaylagawarecki Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

janeyx99 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

danielvegamyhre Jan 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

danielvegamyhre Jan 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mikaylagawarecki Jan 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

janeyx99 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

danielvegamyhre commented Jan 26, 2026

Uh oh!

danielvegamyhre commented Jan 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

danielvegamyhre commented Jan 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

andrewor14 commented Jan 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

janeyx99 commented Jan 27, 2026

Uh oh!

andrewor14 commented Jan 28, 2026

Uh oh!

andrewor14 Feb 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

andrewor14 commented Jan 9, 2026 •

edited

Loading

pytorch-bot bot commented Jan 9, 2026 •

edited

Loading

janeyx99 left a comment •

edited

Loading

mikaylagawarecki Jan 9, 2026 •

edited

Loading

danielvegamyhre Jan 15, 2026 •

edited

Loading

danielvegamyhre Jan 16, 2026 •

edited

Loading

mikaylagawarecki Jan 16, 2026 •

edited

Loading

danielvegamyhre commented Jan 26, 2026 •

edited

Loading

danielvegamyhre commented Jan 27, 2026 •

edited

Loading

andrewor14 commented Jan 27, 2026 •

edited

Loading

andrewor14 Feb 6, 2026 •

edited

Loading