pytorch - 💡(How to fix) Fix `AssertionError: expected size/stride mismatch` in Inductor-generated code for `torch.fft.fft` after `MultiheadAttention`, with warning about unsupported complex op codegen [1 participants]

🐛 Describe the bug

torch.compile(backend='inductor') fails at runtime (after successful compilation) with a stride assertion error inside the generated kernel:

assert_size_stride(buf24, (4, 12, 4), (48, 4, 1), 'torch.ops.aten._fft_r2c.default')
AssertionError: expected size 4==4, stride 4==48 at dim=0; expected size 12==12, stride 16==4 at dim=1
Error in op: torch.ops.aten._fft_r2c.default

Inductor allocated a buffer with strides (16, 4, 1) but passed it to _fft_r2c expecting strides (48, 4, 1). This is an incorrect buffer layout computed by Inductor — a mismatched stride means Inductor's size/stride inference for the intermediate buffer feeding fft is wrong.

Prior to the crash, Inductor itself emits a warning:

UserWarning: Torchinductor does not support code generation for complex operators. Performance may be worse than eager.

This warning is directly relevant: torch.fft.fft produces complex output, and Inductor's fallback path for complex ops appears to allocate the pre-FFT real buffer with incorrect strides, causing the downstream _fft_r2c call to fail its size/stride assertion.

Eager mode runs correctly and produces finite outputs.

The model pipeline is:

1. F.scaled_dot_product_attention(x, x, x) on input [4, 12, 4]
2. nn.MultiheadAttention(4, 2, batch_first=True).eval() → [4, 12, 4]
3. torch.fft.fft(...).abs() → [4, 12, 4] (real)
4. nn.Hardswish()

Traceback (abridged)

File ".../output_code.py", line 656, in __call__
    return self.current_callable(inputs)
File "/tmp/inductor_.../[generated].py", line 474, in call
    assert_size_stride(buf24, (4, 12, 4), (48, 4, 1), 'torch.ops.aten._fft_r2c.default')
AssertionError: expected size 4==4, stride 4==48 at dim=0;
               expected size 12==12, stride 16==4 at dim=1
Error in op: torch.ops.aten._fft_r2c.default

Full call chain:

torch.compile()
  → aot_autograd → runtime_wrappers
    → compiled_fn (Inductor-generated .py)
      → assert_size_stride(buf24, ...)  ← crash
          op: torch.ops.aten._fft_r2c.default

Minimal Reproducer

import os, tempfile
os.environ.setdefault('TORCHINDUCTOR_CACHE_DIR', tempfile.mkdtemp(prefix='inductor_'))

import torch
import torch.nn as nn
import torch.nn.functional as F

m_mha      = nn.MultiheadAttention(4, 2, batch_first=True).eval()
m_hardswish = nn.Hardswish()

torch.manual_seed(0)
x = torch.randn([4, 12, 4])

def model():
    out = F.scaled_dot_product_attention(x, x, x)   # [4, 12, 4]
    out, _ = m_mha(out, out, out)                    # [4, 12, 4]
    out = torch.fft.fft(out).abs()                   # [4, 12, 4] real
    out = m_hardswish(out)                           # [4, 12, 4]
    return out

# Eager: works fine
print("Eager output is finite:", torch.isfinite(model()).all().item())

# Compiled: crashes with stride mismatch in generated code
compiled_model = torch.compile(model, backend='inductor')
compiled_model()  # ← AssertionError: expected size/stride mismatch at _fft_r2c

Error logs

/home/.venv/lib/python3.10/site-packages/torch/_inductor/lowering.py:2212: UserWarning: Torchinductor does not support code generation for complex operators. Performance may be worse than eager. warnings.warn( Traceback (most recent call last): File "/home/bugs/crash_ae7fbc36.py", line 31, in <module> _compiled_out = _compiled() File "/home/.venv/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1024, in compile_wrapper return fn(*args, **kwargs) File "/home/bugs/crash_ae7fbc36.py", line 18, in model_10303 def model_10303(): File "/home/.venv/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1263, in _fn return fn(*args, **kwargs) File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/aot_autograd.py", line 1200, in forward return compiled_fn(full_args) File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 566, in runtime_wrapper all_outs = call_func_at_runtime_with_args( File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/utils.py", line 138, in call_func_at_runtime_with_args out = normalize_as_list(f(args)) File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/utils.py", line 105, in g return f(*args) File "/home/.venv/lib/python3.10/site-packages/torch/autograd/function.py", line 596, in apply return super().apply(*args, **kwargs) # type: ignore[misc] File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 2503, in forward fw_outs = call_func_at_runtime_with_args( File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/utils.py", line 138, in call_func_at_runtime_with_args out = normalize_as_list(f(args)) File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 783, in wrapper return compiled_fn(runtime_args) File "/home/.venv/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 1011, in inner_fn outs = compiled_fn(args) File "/home/.venv/lib/python3.10/site-packages/torch/_inductor/output_code.py", line 656, in call return self.current_callable(inputs) File "/tmp/inductor_cw9h964t/yd/cydjvcsxv5zwpez54smxwuxhyud7sr5mgbb4njo2oocc6rp7wrm5.py", line 474, in call assert_size_stride(buf24, (4, 12, 4), (48, 4, 1), 'torch.ops.aten._fft_r2c.default') AssertionError: expected size 4==4, stride 4==48 at dim=0; expected size 12==12, stride 16==4 at dim=1 Error in op: torch.ops.aten._fft_r2c.default This error most often comes from a incorrect fake (aka meta) kernel for a custom op. Use torch.library.opcheck to test your custom op. See https://pytorch.org/docs/stable/library.html#torch.library.opcheck

Versions

Collecting environment information... PyTorch version: 2.11.0+cu130 Is debug build: False CUDA used to build PyTorch: 13.0 ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.5 LTS (x86_64) GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0 Clang version: 15.0.0 ([email protected]:llvm/llvm-project.git 4ba6a9c9f65bbc8bd06e3652cb20fd4dfc846137) CMake version: version 3.22.1 Libc version: glibc-2.35

Python version: 3.10.12 (main, Mar 3 2026, 11:56:32) [GCC 11.4.0] (64-bit runtime) Python platform: Linux-6.8.0-94-generic-x86_64-with-glibc2.35 Is CUDA available: False CUDA runtime version: No CUDA CUDA_MODULE_LOADING set to: N/A GPU models and configuration: No CUDA Nvidia driver version: No CUDA cuDNN version: No CUDA Is XPU available: False HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True Caching allocator config: N/A

CPU: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 52 bits physical, 57 bits virtual Byte Order: Little Endian CPU(s): 384 On-line CPU(s) list: 0-383 Vendor ID: AuthenticAMD Model name: AMD EPYC 9684X 96-Core Processor CPU family: 25 Model: 17 Thread(s) per core: 2 Core(s) per socket: 96 Socket(s): 2 Stepping: 2 BogoMIPS: 5099.98 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local user_shstk avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin cppc amd_ibpb_ret arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid overflow_recov succor smca fsrm flush_l1d debug_swap ibpb_exit_to_user Virtualization: AMD-V L1d cache: 6 MiB (192 instances) L1i cache: 6 MiB (192 instances) L2 cache: 192 MiB (192 instances) L3 cache: 2.3 GiB (24 instances) NUMA node(s): 2 NUMA node0 CPU(s): 0-95,192-287 NUMA node1 CPU(s): 96-191,288-383 Vulnerability Gather data sampling: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling: Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Mitigation; Safe RET Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; STIBP always-on; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Mitigation; IBPB before exit to userspace

Versions of relevant libraries: [pip3] numpy==2.2.6 [pip3] nvidia-cublas==13.1.0.3 [pip3] nvidia-cuda-cupti==13.0.85 [pip3] nvidia-cuda-nvrtc==13.0.88 [pip3] nvidia-cuda-runtime==13.0.96 [pip3] nvidia-cudnn-cu13==9.19.0.56 [pip3] nvidia-cufft==12.0.0.61 [pip3] nvidia-curand==10.4.0.35 [pip3] nvidia-cusolver==12.0.4.66 [pip3] nvidia-cusparse==12.6.3.3 [pip3] nvidia-cusparselt-cu13==0.8.0 [pip3] nvidia-nccl-cu13==2.28.9 [pip3] nvidia-nvjitlink==13.0.88 [pip3] nvidia-nvtx==13.0.85 [pip3] torch==2.11.0 [pip3] triton==3.6.0 [conda] Could not collect

cc @ezyang @anjali411 @dylanbespalko @mruberry @nikitaved @amjames @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @aakhundov @coconutruben @jataylo