Issue Station
Sign In
Tools hereToolbox

pytorch - 💡(How to fix) Fix [Inductor][CUDA] Internal KeyError: 'complex64' during Triton signature generation for graph with complex intermediates

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…

Error Message

import torch

def target_function( l_tok_embeddings_weight_, l_tokens_, l_attn_norm_0_weight_, l_wq_0_weight_, l_wk_0_weight_, l_wv_0_weight_, l_wo_0_weight_ ): arange = torch.arange(0, 8, 2, device='cuda', dtype=torch.float32) angles = torch.true_divide(arange, 8) pow_1 = torch.pow(10000.0, angles) freqs = torch.true_divide(1.0, pow_1) t = torch.arange(0, 4, device='cuda', dtype=torch.float32) outer = torch.outer(t, freqs) freqs_1 = outer.float() ones_like = torch.ones_like(freqs_1) freqs_cis = torch.polar(ones_like, freqs_1)

h = torch.nn.functional.embedding(l_tokens_, l_tok_embeddings_weight_)
x_normed = h.float()
pow_2 = x_normed.pow(2)
mean = pow_2.mean(-1, keepdim=True)
add = torch.add(mean, 1e-05)
rsqrt = torch.rsqrt(add)
x_normed_1 = torch.mul(x_normed, rsqrt)
type_as = x_normed_1.type_as(h)
normed_x = torch.mul(type_as, l_attn_norm_0_weight_)

q = torch.nn.functional.linear(normed_x, l_wq_0_weight_)
k = torch.nn.functional.linear(normed_x, l_wk_0_weight_)
v = torch.nn.functional.linear(normed_x, l_wv_0_weight_)

view = q.view(2, 4, 8, 8)
q_1 = view.transpose(1, 2)
view_1 = k.view(2, 4, 8, 8)
k_1 = view_1.transpose(1, 2)
view_2 = v.view(2, 4, 8, 8)
v_1 = view_2.transpose(1, 2)

float_3 = q_1.float()
reshape = float_3.reshape(2, 8, 4, -1, 2)
xq_ = torch.view_as_complex(reshape)

float_4 = k_1.float()
reshape_1 = float_4.reshape(2, 8, 4, -1, 2)
xk_ = torch.view_as_complex(reshape_1)

freqs_cis_1 = freqs_cis.reshape(1, 1, 4, -1)
mul_2 = torch.mul(xq_, freqs_cis_1)
view_as_real = torch.view_as_real(mul_2)
xq_out = view_as_real.flatten(3)

mul_3 = torch.mul(xk_, freqs_cis_1)
view_as_real_1 = torch.view_as_real(mul_3)
xk_out = view_as_real_1.flatten(3)

q_2 = xq_out.type_as(q_1)
k_2 = xk_out.type_as(k_1)

mask = torch.full([1, 1, 4, 4], float('-inf'), device='cuda', dtype=torch.float32)
mask_1 = torch.triu(mask, diagonal=1)
transpose_3 = k_2.transpose(-2, -1)
matmul = torch.matmul(q_2, transpose_3)

scores = torch.mul(matmul, 0.35355339059327373)
scores_1 = torch.add(scores, mask_1)
attn_weights = torch.nn.functional.softmax(scores_1, dim=-1)
attn_output = torch.matmul(attn_weights, v_1)

transpose_4 = attn_output.transpose(1, 2)
contiguous = transpose_4.contiguous()
attn_output_1 = contiguous.view(2, 4, 64)
linear_3 = torch.nn.functional.linear(attn_output_1, l_wo_0_weight_)

return (arange, t, mask, h, angles, mask_1, x_normed, pow_1, pow_2, freqs, mean, outer, add, freqs_1, rsqrt, ones_like, x_normed_1, freqs_cis, type_as, freqs_cis_1, normed_x, q, k, v, view, view_1, view_2, q_1, k_1, v_1, float_3, float_4, reshape, reshape_1, xq_, xk_, mul_2, mul_3, view_as_real, view_as_real_1, xq_out, xk_out, q_2, k_2, transpose_3, matmul, scores, scores_1, attn_weights, attn_output, transpose_4, contiguous, attn_output_1, linear_3)

def get_inputs(): torch.manual_seed(2077) return ( (torch.randn([512, 64], dtype=torch.float32) * 0.1).cuda(), torch.zeros([2, 4], dtype=torch.int64).cuda(), (torch.randn([64], dtype=torch.float32) * 0.1).cuda(), (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(), (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(), (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(), (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(), )

if name == "main": if not torch.cuda.is_available(): print("CUDA is not available. Please run this script on a GPU node.") exit(1)

inputs = get_inputs()

print("Running torch.compile(mode='autotune') on GPU...")
opt_fn = torch.compile(target_function, mode="max-autotune")

try:
    opt_fn(*inputs)
    print("Compilation Passed (Bug not triggered).")
except Exception as e:
    import traceback
    traceback.print_exc()
    print(f"\n[BINGO!] Crashed with: {type(e).__name__}: {e}")

Root Cause

The CPU variant of the same high-level RoPE graph fails in cpp_bmm_template with an AssertionError. I filed that separately because it goes through a different backend/codegen path.

Fix Action

Fix / Workaround

Error Log

(torch-nightly) xyt19@Oasis:/tmp$ TORCHDYNAMO_VERBOSE=1 python bug.py
Running torch.compile(mode='autotune') on GPU...
/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/lowering.py:2504: UserWarning: Torchinductor does not support code generation for complex operators. Performance may be worse than eager.
  warnings.warn(
/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:320: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
  warnings.warn(
W0528 00:28:45.226000 12081 site-packages/torch/_inductor/utils.py:1884] [0/0] Not enough SMs to use max_autotune_gemm mode
Traceback (most recent call last):
  File "/tmp/bug.py", line 99, in <module>
    opt_fn(*inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1158, in compile_wrapper
    raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1141, in compile_wrapper
    result = fn(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2619, in __call__
    result = self._torchdynamo_orig_backend(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2310, in __call__
    result = self._inner_convert(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 777, in __call__
    result = _compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2094, in _compile
    guarded_code, tracer_output = compile_inner(code, one_graph, hooks)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_utils_internal.py", line 96, in wrapper_function
    return function(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1679, in compile_inner
    result = _compile_inner(code, one_graph, hooks)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1739, in _compile_inner
    dynamo_output = compile_frame(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1584, in compile_frame
    bytecode, tracer_output = transform_code_object(code, transform)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/bytecode_transformation.py", line 1766, in transform_code_object
    tracer_output = transformations(instructions, code_options)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1555, in transform
    tracer_output = trace_frame(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 368, in _fn
    return fn(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 954, in trace_frame
    run_tracer()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 935, in run_tracer
    tracer.run()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 1883, in run
    while self.step():
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 1536, in step
    self.dispatch_table[inst.opcode](self, inst)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 5449, in RETURN_VALUE
    self._return(inst)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 5422, in _return
    all_stack_locals_metadata = self.output.compile_subgraph(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2171, in compile_subgraph
    instructions, subgraph_pycode = self.compile_and_call_fx_graph(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2817, in compile_and_call_fx_graph
    compiled_fn = self.call_user_compiler(gm, self.example_inputs())
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2987, in call_user_compiler
    return self._call_user_compiler(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 3049, in _call_user_compiler
    compiled_fn = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/repro/after_dynamo.py", line 159, in __call__
    compiled_gm = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/__init__.py", line 2482, in __call__
    return compile_fx(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2705, in compile_fx
    return compile_fx(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2764, in compile_fx
    return _maybe_wrap_and_compile_fx_main(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2845, in _maybe_wrap_and_compile_fx_main
    return _compile_fx_main(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 3058, in _compile_fx_main
    raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 3043, in _compile_fx_main
    return dynamo_common.aot_autograd(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/backends/common.py", line 123, in __call__
    cg = aot_module_simplified(gm, example_inputs, **self.kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/aot_autograd.py", line 1238, in aot_module_simplified
    compiled_fn, _ = aot_stage2_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 357, in aot_stage2_compile
    return aot_stage2_inference(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 481, in aot_stage2_inference
    compiled_fw = _aot_stage2b_inference_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 408, in _aot_stage2b_inference_compile
    return _aot_stage2b_compile_forward_or_inference(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 2779, in _aot_stage2b_compile_forward_or_inference
    compiled_fw_func = compiler(fw_module, adjusted_flat_args)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/schemas.py", line 1460, in __call__
    output_code = self.compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2906, in fw_compiler_base
    return compile_fx_forward(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2534, in compile_fx_forward
    result = inner_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/contextlib.py", line 79, in inner
    return func(*args, **kwds)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 836, in compile_fx_inner
    return wrap_compiler_debug(_compile_fx_inner, compiler_name="inductor")(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/repro/after_aot.py", line 317, in debug_wrapper
    inner_compiled_fn = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1078, in _compile_fx_inner
    raise InductorError(e, currentframe()).with_traceback(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1058, in _compile_fx_inner
    mb_compiled_graph = fx_codegen_and_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1845, in fx_codegen_and_compile
    return scheme.codegen_and_compile(gm, example_inputs, inputs_to_check, graph_kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1606, in codegen_and_compile
    compiled_module = graph.compile_to_module()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2661, in compile_to_module
    return self._compile_to_module()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2667, in _compile_to_module
    self.codegen_with_cpp_wrapper() if self.cpp_wrapper else self.codegen()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2603, in codegen
    self.scheduler.codegen()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 8941, in codegen
    self._codegen_partitions()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 9083, in _codegen_partitions
    self._codegen_partition_wrapper(partition, signature)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 8964, in _codegen_partition_wrapper
    self._codegen(partition)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 9235, in _codegen
    self.get_backend(device).codegen_node(node)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/cuda_combined_scheduling.py", line 152, in codegen_node
    return self._triton_scheduling.codegen_node(node)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 1979, in codegen_node
    return self._codegen_nodes(nodes, coalesce_analysis)  # type: ignore[arg-type]
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 1952, in _codegen_nodes
    return self.codegen_node_schedule(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 2076, in codegen_node_schedule
    src_code = kernel.codegen_kernel()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton.py", line 5958, in codegen_kernel
    triton_meta_signature = signature_to_meta(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 159, in signature_to_meta
    return {
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 160, in <dictcomp>
    argdefs[i].name: signature_of(arg, size_dtype=_decide_tl_dtype(arg))
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 42, in signature_of
    typ = _type_of(arg.dtype)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/utils.py", line 534, in _type_of
    return key if isinstance(key, str) else f"*{tys[dtype_str]}"
torch._inductor.exc.InductorError: KeyError: 'complex64'

Code Example

import torch

def target_function(
    l_tok_embeddings_weight_, l_tokens_, l_attn_norm_0_weight_,
    l_wq_0_weight_, l_wk_0_weight_, l_wv_0_weight_, l_wo_0_weight_
):
    arange = torch.arange(0, 8, 2, device='cuda', dtype=torch.float32)
    angles = torch.true_divide(arange, 8)
    pow_1 = torch.pow(10000.0, angles)
    freqs = torch.true_divide(1.0, pow_1)
    t = torch.arange(0, 4, device='cuda', dtype=torch.float32)
    outer = torch.outer(t, freqs)
    freqs_1 = outer.float()
    ones_like = torch.ones_like(freqs_1)
    freqs_cis = torch.polar(ones_like, freqs_1)

    h = torch.nn.functional.embedding(l_tokens_, l_tok_embeddings_weight_)
    x_normed = h.float()
    pow_2 = x_normed.pow(2)
    mean = pow_2.mean(-1, keepdim=True)
    add = torch.add(mean, 1e-05)
    rsqrt = torch.rsqrt(add)
    x_normed_1 = torch.mul(x_normed, rsqrt)
    type_as = x_normed_1.type_as(h)
    normed_x = torch.mul(type_as, l_attn_norm_0_weight_)

    q = torch.nn.functional.linear(normed_x, l_wq_0_weight_)
    k = torch.nn.functional.linear(normed_x, l_wk_0_weight_)
    v = torch.nn.functional.linear(normed_x, l_wv_0_weight_)

    view = q.view(2, 4, 8, 8)
    q_1 = view.transpose(1, 2)
    view_1 = k.view(2, 4, 8, 8)
    k_1 = view_1.transpose(1, 2)
    view_2 = v.view(2, 4, 8, 8)
    v_1 = view_2.transpose(1, 2)

    float_3 = q_1.float()
    reshape = float_3.reshape(2, 8, 4, -1, 2)
    xq_ = torch.view_as_complex(reshape)

    float_4 = k_1.float()
    reshape_1 = float_4.reshape(2, 8, 4, -1, 2)
    xk_ = torch.view_as_complex(reshape_1)

    freqs_cis_1 = freqs_cis.reshape(1, 1, 4, -1)
    mul_2 = torch.mul(xq_, freqs_cis_1)
    view_as_real = torch.view_as_real(mul_2)
    xq_out = view_as_real.flatten(3)

    mul_3 = torch.mul(xk_, freqs_cis_1)
    view_as_real_1 = torch.view_as_real(mul_3)
    xk_out = view_as_real_1.flatten(3)

    q_2 = xq_out.type_as(q_1)
    k_2 = xk_out.type_as(k_1)

    mask = torch.full([1, 1, 4, 4], float('-inf'), device='cuda', dtype=torch.float32)
    mask_1 = torch.triu(mask, diagonal=1)
    transpose_3 = k_2.transpose(-2, -1)
    matmul = torch.matmul(q_2, transpose_3)

    scores = torch.mul(matmul, 0.35355339059327373)
    scores_1 = torch.add(scores, mask_1)
    attn_weights = torch.nn.functional.softmax(scores_1, dim=-1)
    attn_output = torch.matmul(attn_weights, v_1)

    transpose_4 = attn_output.transpose(1, 2)
    contiguous = transpose_4.contiguous()
    attn_output_1 = contiguous.view(2, 4, 64)
    linear_3 = torch.nn.functional.linear(attn_output_1, l_wo_0_weight_)

    return (arange, t, mask, h, angles, mask_1, x_normed, pow_1, pow_2, freqs, mean, outer, add, freqs_1, rsqrt, ones_like, x_normed_1, freqs_cis, type_as, freqs_cis_1, normed_x, q, k, v, view, view_1, view_2, q_1, k_1, v_1, float_3, float_4, reshape, reshape_1, xq_, xk_, mul_2, mul_3, view_as_real, view_as_real_1, xq_out, xk_out, q_2, k_2, transpose_3, matmul, scores, scores_1, attn_weights, attn_output, transpose_4, contiguous, attn_output_1, linear_3)


def get_inputs():
    torch.manual_seed(2077)
    return (
        (torch.randn([512, 64], dtype=torch.float32) * 0.1).cuda(),
        torch.zeros([2, 4], dtype=torch.int64).cuda(),
        (torch.randn([64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
    )

if __name__ == "__main__":
    if not torch.cuda.is_available():
        print("CUDA is not available. Please run this script on a GPU node.")
        exit(1)

    inputs = get_inputs()

    print("Running torch.compile(mode='autotune') on GPU...")
    opt_fn = torch.compile(target_function, mode="max-autotune")

    try:
        opt_fn(*inputs)
        print("Compilation Passed (Bug not triggered).")
    except Exception as e:
        import traceback
        traceback.print_exc()
        print(f"\n[BINGO!] Crashed with: {type(e).__name__}: {e}")

---

(torch-nightly) xyt19@Oasis:/tmp$ TORCHDYNAMO_VERBOSE=1 python bug.py
Running torch.compile(mode='autotune') on GPU...
/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/lowering.py:2504: UserWarning: Torchinductor does not support code generation for complex operators. Performance may be worse than eager.
  warnings.warn(
/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:320: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
  warnings.warn(
W0528 00:28:45.226000 12081 site-packages/torch/_inductor/utils.py:1884] [0/0] Not enough SMs to use max_autotune_gemm mode
Traceback (most recent call last):
  File "/tmp/bug.py", line 99, in <module>
    opt_fn(*inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1158, in compile_wrapper
    raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1141, in compile_wrapper
    result = fn(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2619, in __call__
    result = self._torchdynamo_orig_backend(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2310, in __call__
    result = self._inner_convert(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 777, in __call__
    result = _compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2094, in _compile
    guarded_code, tracer_output = compile_inner(code, one_graph, hooks)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_utils_internal.py", line 96, in wrapper_function
    return function(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1679, in compile_inner
    result = _compile_inner(code, one_graph, hooks)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1739, in _compile_inner
    dynamo_output = compile_frame(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1584, in compile_frame
    bytecode, tracer_output = transform_code_object(code, transform)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/bytecode_transformation.py", line 1766, in transform_code_object
    tracer_output = transformations(instructions, code_options)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1555, in transform
    tracer_output = trace_frame(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 368, in _fn
    return fn(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 954, in trace_frame
    run_tracer()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 935, in run_tracer
    tracer.run()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 1883, in run
    while self.step():
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 1536, in step
    self.dispatch_table[inst.opcode](self, inst)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 5449, in RETURN_VALUE
    self._return(inst)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 5422, in _return
    all_stack_locals_metadata = self.output.compile_subgraph(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2171, in compile_subgraph
    instructions, subgraph_pycode = self.compile_and_call_fx_graph(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2817, in compile_and_call_fx_graph
    compiled_fn = self.call_user_compiler(gm, self.example_inputs())
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2987, in call_user_compiler
    return self._call_user_compiler(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 3049, in _call_user_compiler
    compiled_fn = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/repro/after_dynamo.py", line 159, in __call__
    compiled_gm = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/__init__.py", line 2482, in __call__
    return compile_fx(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2705, in compile_fx
    return compile_fx(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2764, in compile_fx
    return _maybe_wrap_and_compile_fx_main(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2845, in _maybe_wrap_and_compile_fx_main
    return _compile_fx_main(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 3058, in _compile_fx_main
    raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 3043, in _compile_fx_main
    return dynamo_common.aot_autograd(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/backends/common.py", line 123, in __call__
    cg = aot_module_simplified(gm, example_inputs, **self.kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/aot_autograd.py", line 1238, in aot_module_simplified
    compiled_fn, _ = aot_stage2_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 357, in aot_stage2_compile
    return aot_stage2_inference(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 481, in aot_stage2_inference
    compiled_fw = _aot_stage2b_inference_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 408, in _aot_stage2b_inference_compile
    return _aot_stage2b_compile_forward_or_inference(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 2779, in _aot_stage2b_compile_forward_or_inference
    compiled_fw_func = compiler(fw_module, adjusted_flat_args)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/schemas.py", line 1460, in __call__
    output_code = self.compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2906, in fw_compiler_base
    return compile_fx_forward(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2534, in compile_fx_forward
    result = inner_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/contextlib.py", line 79, in inner
    return func(*args, **kwds)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 836, in compile_fx_inner
    return wrap_compiler_debug(_compile_fx_inner, compiler_name="inductor")(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/repro/after_aot.py", line 317, in debug_wrapper
    inner_compiled_fn = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1078, in _compile_fx_inner
    raise InductorError(e, currentframe()).with_traceback(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1058, in _compile_fx_inner
    mb_compiled_graph = fx_codegen_and_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1845, in fx_codegen_and_compile
    return scheme.codegen_and_compile(gm, example_inputs, inputs_to_check, graph_kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1606, in codegen_and_compile
    compiled_module = graph.compile_to_module()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2661, in compile_to_module
    return self._compile_to_module()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2667, in _compile_to_module
    self.codegen_with_cpp_wrapper() if self.cpp_wrapper else self.codegen()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2603, in codegen
    self.scheduler.codegen()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 8941, in codegen
    self._codegen_partitions()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 9083, in _codegen_partitions
    self._codegen_partition_wrapper(partition, signature)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 8964, in _codegen_partition_wrapper
    self._codegen(partition)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 9235, in _codegen
    self.get_backend(device).codegen_node(node)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/cuda_combined_scheduling.py", line 152, in codegen_node
    return self._triton_scheduling.codegen_node(node)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 1979, in codegen_node
    return self._codegen_nodes(nodes, coalesce_analysis)  # type: ignore[arg-type]
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 1952, in _codegen_nodes
    return self.codegen_node_schedule(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 2076, in codegen_node_schedule
    src_code = kernel.codegen_kernel()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton.py", line 5958, in codegen_kernel
    triton_meta_signature = signature_to_meta(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 159, in signature_to_meta
    return {
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 160, in <dictcomp>
    argdefs[i].name: signature_of(arg, size_dtype=_decide_tl_dtype(arg))
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 42, in signature_of
    typ = _type_of(arg.dtype)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/utils.py", line 534, in _type_of
    return key if isinstance(key, str) else f"*{tys[dtype_str]}"
torch._inductor.exc.InductorError: KeyError: 'complex64'


[BINGO!] Crashed with: InductorError: KeyError: 'complex64'
RAW_BUFFERClick to expand / collapse

🐛 Describe the bug

When compiling a function that uses complex number operations (e.g., torch.polar, torch.view_as_complex) with torch.compile(mode="max-autotune"), the compiler crashes with a KeyError: 'complex64'.

Although Inductor correctly issues a warning that it does not support code generation for complex operators, it fails to gracefully fall back to eager mode. Instead, it crashes internally during Triton kernel signature generation inside torch/_inductor/utils.py.

The CPU variant of the same high-level RoPE graph fails in cpp_bmm_template with an AssertionError. I filed that separately because it goes through a different backend/codegen path.

To Reproduce

Run the following standalone script:

import torch

def target_function(
    l_tok_embeddings_weight_, l_tokens_, l_attn_norm_0_weight_,
    l_wq_0_weight_, l_wk_0_weight_, l_wv_0_weight_, l_wo_0_weight_
):
    arange = torch.arange(0, 8, 2, device='cuda', dtype=torch.float32)
    angles = torch.true_divide(arange, 8)
    pow_1 = torch.pow(10000.0, angles)
    freqs = torch.true_divide(1.0, pow_1)
    t = torch.arange(0, 4, device='cuda', dtype=torch.float32)
    outer = torch.outer(t, freqs)
    freqs_1 = outer.float()
    ones_like = torch.ones_like(freqs_1)
    freqs_cis = torch.polar(ones_like, freqs_1)

    h = torch.nn.functional.embedding(l_tokens_, l_tok_embeddings_weight_)
    x_normed = h.float()
    pow_2 = x_normed.pow(2)
    mean = pow_2.mean(-1, keepdim=True)
    add = torch.add(mean, 1e-05)
    rsqrt = torch.rsqrt(add)
    x_normed_1 = torch.mul(x_normed, rsqrt)
    type_as = x_normed_1.type_as(h)
    normed_x = torch.mul(type_as, l_attn_norm_0_weight_)

    q = torch.nn.functional.linear(normed_x, l_wq_0_weight_)
    k = torch.nn.functional.linear(normed_x, l_wk_0_weight_)
    v = torch.nn.functional.linear(normed_x, l_wv_0_weight_)

    view = q.view(2, 4, 8, 8)
    q_1 = view.transpose(1, 2)
    view_1 = k.view(2, 4, 8, 8)
    k_1 = view_1.transpose(1, 2)
    view_2 = v.view(2, 4, 8, 8)
    v_1 = view_2.transpose(1, 2)

    float_3 = q_1.float()
    reshape = float_3.reshape(2, 8, 4, -1, 2)
    xq_ = torch.view_as_complex(reshape)

    float_4 = k_1.float()
    reshape_1 = float_4.reshape(2, 8, 4, -1, 2)
    xk_ = torch.view_as_complex(reshape_1)

    freqs_cis_1 = freqs_cis.reshape(1, 1, 4, -1)
    mul_2 = torch.mul(xq_, freqs_cis_1)
    view_as_real = torch.view_as_real(mul_2)
    xq_out = view_as_real.flatten(3)

    mul_3 = torch.mul(xk_, freqs_cis_1)
    view_as_real_1 = torch.view_as_real(mul_3)
    xk_out = view_as_real_1.flatten(3)

    q_2 = xq_out.type_as(q_1)
    k_2 = xk_out.type_as(k_1)

    mask = torch.full([1, 1, 4, 4], float('-inf'), device='cuda', dtype=torch.float32)
    mask_1 = torch.triu(mask, diagonal=1)
    transpose_3 = k_2.transpose(-2, -1)
    matmul = torch.matmul(q_2, transpose_3)

    scores = torch.mul(matmul, 0.35355339059327373)
    scores_1 = torch.add(scores, mask_1)
    attn_weights = torch.nn.functional.softmax(scores_1, dim=-1)
    attn_output = torch.matmul(attn_weights, v_1)

    transpose_4 = attn_output.transpose(1, 2)
    contiguous = transpose_4.contiguous()
    attn_output_1 = contiguous.view(2, 4, 64)
    linear_3 = torch.nn.functional.linear(attn_output_1, l_wo_0_weight_)

    return (arange, t, mask, h, angles, mask_1, x_normed, pow_1, pow_2, freqs, mean, outer, add, freqs_1, rsqrt, ones_like, x_normed_1, freqs_cis, type_as, freqs_cis_1, normed_x, q, k, v, view, view_1, view_2, q_1, k_1, v_1, float_3, float_4, reshape, reshape_1, xq_, xk_, mul_2, mul_3, view_as_real, view_as_real_1, xq_out, xk_out, q_2, k_2, transpose_3, matmul, scores, scores_1, attn_weights, attn_output, transpose_4, contiguous, attn_output_1, linear_3)


def get_inputs():
    torch.manual_seed(2077)
    return (
        (torch.randn([512, 64], dtype=torch.float32) * 0.1).cuda(),
        torch.zeros([2, 4], dtype=torch.int64).cuda(),
        (torch.randn([64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
        (torch.randn([64, 64], dtype=torch.float32) * 0.1).cuda(),
    )

if __name__ == "__main__":
    if not torch.cuda.is_available():
        print("CUDA is not available. Please run this script on a GPU node.")
        exit(1)

    inputs = get_inputs()

    print("Running torch.compile(mode='autotune') on GPU...")
    opt_fn = torch.compile(target_function, mode="max-autotune")

    try:
        opt_fn(*inputs)
        print("Compilation Passed (Bug not triggered).")
    except Exception as e:
        import traceback
        traceback.print_exc()
        print(f"\n[BINGO!] Crashed with: {type(e).__name__}: {e}")

Expected behavior I am not requesting full complex operator codegen support or optimized lowering for this graph. If complex tensors/operators are currently unsupported by Inductor/Triton codegen, torch.compile should handle this gracefully, for example by graph-breaking, falling back where possible, or raising a clear user-facing "unsupported complex dtype/operator" diagnostic.

Error Log

(torch-nightly) xyt19@Oasis:/tmp$ TORCHDYNAMO_VERBOSE=1 python bug.py
Running torch.compile(mode='autotune') on GPU...
/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/lowering.py:2504: UserWarning: Torchinductor does not support code generation for complex operators. Performance may be worse than eager.
  warnings.warn(
/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:320: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
  warnings.warn(
W0528 00:28:45.226000 12081 site-packages/torch/_inductor/utils.py:1884] [0/0] Not enough SMs to use max_autotune_gemm mode
Traceback (most recent call last):
  File "/tmp/bug.py", line 99, in <module>
    opt_fn(*inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1158, in compile_wrapper
    raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 1141, in compile_wrapper
    result = fn(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2619, in __call__
    result = self._torchdynamo_orig_backend(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2310, in __call__
    result = self._inner_convert(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 777, in __call__
    result = _compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 2094, in _compile
    guarded_code, tracer_output = compile_inner(code, one_graph, hooks)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_utils_internal.py", line 96, in wrapper_function
    return function(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1679, in compile_inner
    result = _compile_inner(code, one_graph, hooks)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1739, in _compile_inner
    dynamo_output = compile_frame(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1584, in compile_frame
    bytecode, tracer_output = transform_code_object(code, transform)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/bytecode_transformation.py", line 1766, in transform_code_object
    tracer_output = transformations(instructions, code_options)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 1555, in transform
    tracer_output = trace_frame(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 368, in _fn
    return fn(*args, **kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 954, in trace_frame
    run_tracer()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 935, in run_tracer
    tracer.run()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 1883, in run
    while self.step():
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 1536, in step
    self.dispatch_table[inst.opcode](self, inst)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 5449, in RETURN_VALUE
    self._return(inst)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 5422, in _return
    all_stack_locals_metadata = self.output.compile_subgraph(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2171, in compile_subgraph
    instructions, subgraph_pycode = self.compile_and_call_fx_graph(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2817, in compile_and_call_fx_graph
    compiled_fn = self.call_user_compiler(gm, self.example_inputs())
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 2987, in call_user_compiler
    return self._call_user_compiler(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 3049, in _call_user_compiler
    compiled_fn = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/repro/after_dynamo.py", line 159, in __call__
    compiled_gm = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/__init__.py", line 2482, in __call__
    return compile_fx(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2705, in compile_fx
    return compile_fx(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2764, in compile_fx
    return _maybe_wrap_and_compile_fx_main(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2845, in _maybe_wrap_and_compile_fx_main
    return _compile_fx_main(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 3058, in _compile_fx_main
    raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 3043, in _compile_fx_main
    return dynamo_common.aot_autograd(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/backends/common.py", line 123, in __call__
    cg = aot_module_simplified(gm, example_inputs, **self.kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/aot_autograd.py", line 1238, in aot_module_simplified
    compiled_fn, _ = aot_stage2_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 357, in aot_stage2_compile
    return aot_stage2_inference(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 481, in aot_stage2_inference
    compiled_fw = _aot_stage2b_inference_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 408, in _aot_stage2b_inference_compile
    return _aot_stage2b_compile_forward_or_inference(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/graph_compile.py", line 2779, in _aot_stage2b_compile_forward_or_inference
    compiled_fw_func = compiler(fw_module, adjusted_flat_args)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_functorch/_aot_autograd/schemas.py", line 1460, in __call__
    output_code = self.compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2906, in fw_compiler_base
    return compile_fx_forward(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 2534, in compile_fx_forward
    result = inner_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/contextlib.py", line 79, in inner
    return func(*args, **kwds)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 836, in compile_fx_inner
    return wrap_compiler_debug(_compile_fx_inner, compiler_name="inductor")(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_dynamo/repro/after_aot.py", line 317, in debug_wrapper
    inner_compiled_fn = compiler_fn(gm, example_inputs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1078, in _compile_fx_inner
    raise InductorError(e, currentframe()).with_traceback(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1058, in _compile_fx_inner
    mb_compiled_graph = fx_codegen_and_compile(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1845, in fx_codegen_and_compile
    return scheme.codegen_and_compile(gm, example_inputs, inputs_to_check, graph_kwargs)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 1606, in codegen_and_compile
    compiled_module = graph.compile_to_module()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2661, in compile_to_module
    return self._compile_to_module()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2667, in _compile_to_module
    self.codegen_with_cpp_wrapper() if self.cpp_wrapper else self.codegen()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/graph.py", line 2603, in codegen
    self.scheduler.codegen()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 8941, in codegen
    self._codegen_partitions()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 9083, in _codegen_partitions
    self._codegen_partition_wrapper(partition, signature)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 8964, in _codegen_partition_wrapper
    self._codegen(partition)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/scheduler.py", line 9235, in _codegen
    self.get_backend(device).codegen_node(node)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/cuda_combined_scheduling.py", line 152, in codegen_node
    return self._triton_scheduling.codegen_node(node)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 1979, in codegen_node
    return self._codegen_nodes(nodes, coalesce_analysis)  # type: ignore[arg-type]
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 1952, in _codegen_nodes
    return self.codegen_node_schedule(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/simd.py", line 2076, in codegen_node_schedule
    src_code = kernel.codegen_kernel()
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton.py", line 5958, in codegen_kernel
    triton_meta_signature = signature_to_meta(
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 159, in signature_to_meta
    return {
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 160, in <dictcomp>
    argdefs[i].name: signature_of(arg, size_dtype=_decide_tl_dtype(arg))
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/codegen/triton_utils.py", line 42, in signature_of
    typ = _type_of(arg.dtype)
  File "/home/xyt19/miniconda3/envs/torch-nightly/lib/python3.10/site-packages/torch/_inductor/utils.py", line 534, in _type_of
    return key if isinstance(key, str) else f"*{tys[dtype_str]}"
torch._inductor.exc.InductorError: KeyError: 'complex64'


[BINGO!] Crashed with: InductorError: KeyError: 'complex64'

Versions

PyTorch version: 2.13.0.dev20260521+cu130 Is debug build: False CUDA used to build PyTorch: 13.0 ROCM used to build PyTorch: N/A

OS: Ubuntu 24.04.4 LTS (x86_64) GCC version: (Ubuntu 13.3.0-6ubuntu2~24.04.1) 13.3.0 Clang version: 18.1.3 (1ubuntu1) CMake version: version 3.28.3 Libc version: glibc-2.39

Python version: 3.10.20 (main, Mar 11 2026, 17:46:40) [GCC 14.3.0] (64-bit runtime) Python platform: Linux-5.15.167.4-microsoft-standard-WSL2-x86_64-with-glibc2.39 Is CUDA available: True CUDA runtime version: 12.0.140 Nvidia driver version: 596.49 cuDNN version: Probably one of the following: /usr/lib/x86_64-linux-gnu/libcudnn.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_engines_tensor_ir.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.21.1 /usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.21.1 Is XPU available: False HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True Caching allocator config: N/A ersions of relevant libraries: [pip3] numpy==2.2.6 [pip3] nvidia-cublas==13.1.1.3 [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==9.20.0.48 [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==0.8.1 [pip3] nvidia-nccl-cu13==2.29.7 [pip3] nvidia-nvjitlink==13.0.88 [pip3] nvidia-nvtx==13.0.85 [pip3] torch==2.13.0.dev20260521+cu130 [pip3] torchaudio==2.11.0.dev20260525+cu130 [pip3] torchvision==0.28.0.dev20260525+cu130 [pip3] triton==3.7.0+git88b227e2 [conda] numpy 2.2.6 pypi_0 pypi [conda] nvidia-cublas 13.1.1.3 pypi_0 pypi [conda] nvidia-cuda-cupti 13.0.85 pypi_0 pypi [conda] nvidia-cuda-nvrtc 13.0.88 pypi_0 pypi [conda] nvidia-cuda-runtime 13.0.96 pypi_0 pypi [conda] nvidia-cudnn-cu13 9.20.0.48 pypi_0 pypi [conda] nvidia-cufft 12.0.0.61 pypi_0 pypi [conda] nvidia-curand 10.4.0.35 pypi_0 pypi [conda] nvidia-cusolver 12.0.4.66 pypi_0 pypi [conda] nvidia-cusparse 12.6.3.3 pypi_0 pypi [conda] nvidia-cusparselt-cu13 0.8.1 pypi_0 pypi [conda] nvidia-nccl-cu13 2.29.7 pypi_0 pypi [conda] nvidia-nvjitlink 13.0.88 pypi_0 pypi [conda] nvidia-nvtx 13.0.85 pypi_0 pypi [conda] torch 2.13.0.dev20260521+cu130 pypi_0 pypi [conda] torchaudio 2.11.0.dev20260525+cu130 pypi_0 pypi [conda] torchvision 0.28.0.dev20260525+cu130 pypi_0 pypi [conda] triton 3.7.0+git88b227e2 pypi_0 pypi

cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @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

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

pytorch - 💡(How to fix) Fix [Inductor][CUDA] Internal KeyError: 'complex64' during Triton signature generation for graph with complex intermediates