Below are all known graph breaks detected by Dynamo.

• GB0000 — All torch_function overrides returned NotImplemented due to TypeError from user code
• GB0001 — Argument of as_subclass must be a non-dispatcher-style tensor subclass
• GB0002 — Assertion failed on symbolic shapes
• GB0003 — Attempt to trace generator
• GB0004 — Attempted super().delattr() on an object without mutation tracking
• GB0005 — Attempted to a str() method implemented in C/C++
• GB0006 — Attempted to call a super() attribute that is not a function or method
• GB0007 — Attempted to call function marked as skipped
• GB0008 — Attempted to inline function marked as skipped
• GB0009 — Attempted to inline function marked as skipped (SkipFunctionVariable)
• GB0010 — Attempted to read a deleted variable
• GB0011 — Attempted to read undefined local variable
• GB0012 — Attempted to read undefined local variable (implicit)
• GB0013 — Attempted to represent unregistered RemovableHandle
• GB0014 — Attempted to wrap RNN, GRU, or LSTM
• GB0015 — Attempted to wrap sparse Tensor
• GB0016 — Attempted to wrap strided NestedTensor
• GB0017 — Attempted to wrap torch._higher_order_ops.invoke_subgraph
• GB0018 — Attempted to wrap unbacked SymInt
• GB0019 — AutogradFunctionContextVariable escaped Dynamo-traced region
• GB0020 — BUILD_STRING key conflict
• GB0021 — BUILD_STRING type error
• GB0022 — Bad import result
• GB0023 — Builtin operator.* comparison with constant self failed
• GB0024 — CLEANUP_THROW with StopIteration
• GB0025 — Call to torch._dynamo.graph_break()
• GB0026 — Calling subclass default constructor with more than tensor argument
• GB0027 — Cannot check Tensor object identity without its fake value
• GB0028 — Caught non-Exception value
• GB0029 — Compilation of intermediate hooks requires compiled autograd
• GB0030 — ComptimeContext graph break
• GB0031 — Custom getattribute in nn.Module attribute access
• GB0032 — Custom getattribute in nn.Module dict key check
• GB0033 — Data dependent operator
• GB0034 — Data-dependent assertion failed (cannot compile partial graph)
• GB0035 — Data-dependent branching with non-constant bool
• GB0036 — Dynamic shape operator
• GB0037 — Dynamic shape operator (no meta kernel)
• GB0038 — Dynamic slicing with Tensor arguments
• GB0039 — Dynamo cache limit exceeded
• GB0040 — Encountered aliasing during higher order op tracing
• GB0041 — Encountered input mutation during higher order op tracing
• GB0042 — Encountered non user function variable during invoke_subgraph HOP tracing
• GB0043 — Encountered non-PT2-compliant op
• GB0044 — Encountered strided NestedTensor in automatic dynamic dim determination
• GB0045 — Encountered tensor.is_inference() during tracing
• GB0046 — Encountered torch.is_inference_mode_enabled during tracing
• GB0047 — Encountered unconverted argument when attempting to inline
• GB0048 — Error getting associated real value
• GB0049 — Error when attempting to resolve op packet
• GB0050 — Exception with bad expected type
• GB0051 — Exception with non-type expectation
• GB0052 — Excessive RestartAnalysis() calls
• GB0053 — FSDP with use_orig_params=False
• GB0054 — Failed to construct Enum variable
• GB0055 — Failed to convert args/kwargs to proxy
• GB0056 — Failed to mutate tensor data attribute
• GB0057 — Failed to raise exception
• GB0058 — Failed to set tensor attribute
• GB0059 — Failed to trace builtin operator
• GB0060 — Failed to trace unittest method
• GB0061 — Failed to unpack object for BUILD_LIST_UNPACK
• GB0062 — Failed to unpack object for UNPACK_EX
• GB0063 — Failed to unpack object for UNPACK_SEQUENCE
• GB0064 — Fake tensor propagation exception
• GB0065 — Graph break in inlined function
• GB0066 — Graph break under GenericContextWrappingVariable
• GB0067 — HigherOrderOperator: Mutating a variable not in the current scope (SideEffects)
• GB0068 — Illegal method invocation in strict mode
• GB0069 — Import failure
• GB0070 — Indexing list with non-scalar tensor
• GB0071 — Inline attempt with self
• GB0072 — Inplace op on input tensor
• GB0073 — Invoking an nn.Module inside a HigherOrderOperator
• GB0074 — Invoking an nn.Module inside a higher order operator
• GB0075 — LOAD_BUILD_CLASS bytecode not supported
• GB0076 — LOAD_FAST_CHECK on uninitialized variable
• GB0077 — Length mismatch when unpacking object for UNPACK_SEQUENCE
• GB0078 — Limitation of `nonstrict_trace
• GB0079 — Missing CALL_INTRINSIC_1 handler
• GB0080 — Missing FakeTensor example value
• GB0081 — Missing attribute when running call_method node
• GB0082 — Missing bytecode handler
• GB0083 — Module-level backwards hooks require compiled autograd.
• GB0084 — Non-constant attribute given to super().__delattr__()
• GB0085 — Non-function or method in subclass of torch.autograd.Function
• GB0086 — Not a Python constant
• GB0087 — NotImplementedError/UnsupportedFakeTensorException when running FX node
• GB0088 — Observed exception
• GB0089 — Observed exception (EXCEPT_HANDLER)
• GB0090 — Operator does not support running with fake tensors
• GB0091 — Read uninitialized cell
• GB0092 — Reconstruction failure
• GB0093 — Reconstruction failure: source.reconstruct not implemented
• GB0094 — SEND with bad type
• GB0095 — Set Exception object __traceback__ attribute to not-None
• GB0096 — Should not compile partial graph (STORE_ATTR)
• GB0097 — Side effect on existing deque with limited maxlen
• GB0098 — Skip calling torch.compiler.disable()d function
• GB0099 — Skip inlining torch.compiler.disable()d function
• GB0100 — Storing Tensor hook handle in globals
• GB0101 — Storing Tensor hook handle in globals (inline call)
• GB0102 — Strict mode banned op
• GB0103 — Tensor subclass overridden method call
• GB0104 — Tensor with grad_fn()
• GB0105 — Tensor.numpy() with trace_numpy=False
• GB0106 — Tensor.numpy() without NumPy installed
• GB0107 — Tensor.random_ op
• GB0108 — Tensor.retain_grad() with AOTDispatcher
• GB0109 — Tensor.tolist() with non-integer tensor
• GB0110 — Tensor.uniform_ op called with from keyword
• GB0111 — TypeError from user code
• GB0112 — TypeError when making fake tensor call
• GB0113 — Unable to resolve super getattr
• GB0114 — Unexpected failure during itertools.accumulate() iteration
• GB0115 — Unexpected failure during itertools.groupby() iteration
• GB0116 — Unexpected type in sourceless builder
• GB0117 — Unhandled args for method
• GB0118 — Unimplemented next() call
• GB0119 — Uninitialized nn.Module
• GB0120 — Unreachable sub-generator code
• GB0121 — UnspecializedNNModuleVariable missing method
• GB0122 — Unsupported SourceType
• GB0123 — Unsupported Tensor.backward() call
• GB0124 — Unsupported Tensor.item() call with capture_scalar_outputs=False
• GB0125 — Unsupported Tensor.requires_grad_() call
• GB0126 — Unsupported Tensor.resize_() call
• GB0127 — Unsupported Tensor.resize_as_() call
• GB0128 — Unsupported Tensor.set_() call
• GB0129 — Unsupported Tensor.sparse_resize_() call
• GB0130 — Unsupported Tensor.sparse_resize_and_clear_() call
• GB0131 — Unsupported setitem/setattr inline attempt
• GB0132 — Unsupported func in itertools.accumulate
• GB0133 — Unsupported arguments for itertools.accumulate
• GB0134 — Unsupported arguments for itertools.groupby
• GB0135 — Unsupported attribute assignment on Exception object
• GB0136 — Unsupported attribute for range() object
• GB0137 — Unsupported attribute for slice() object
• GB0138 — Unsupported autograd.Function context save_for_backward
• GB0139 — Unsupported autograd.Function context method
• GB0140 — Unsupported autograd.Function method
• GB0141 — Unsupported call_id() without source
• GB0142 — Unsupported context manager
• GB0143 — Unsupported conversion for slice assignment
• GB0144 — Unsupported custom jvp
• GB0145 — Unsupported custom vjp
• GB0146 — Unsupported event method
• GB0147 — Unsupported function call
• GB0148 — Unsupported function call (delayed)
• GB0149 — Unsupported functorch tracing attempt
• GB0150 — Unsupported hasattr call
• GB0151 — Unsupported inspect call
• GB0152 — Unsupported key type for itertools.groupby
• GB0153 — Unsupported key type for nn.Module.getitem
• GB0154 — Unsupported kwargs for itertools.accumulate
• GB0155 — Unsupported kwargs for itertools.groupby
• GB0156 — Unsupported method call
• GB0157 — Unsupported ndarray attribute access
• GB0158 — Unsupported ndarray method call
• GB0159 — Unsupported ndarray.version access
• GB0160 — Unsupported next() call
• GB0161 — Unsupported nn.Module attribute type
• GB0162 — Unsupported super().init() call
• GB0163 — Unsupported tensor subclass attribute access
• GB0164 — Unsupported tensor subclass overridden attribute access
• GB0165 — Unsupported torch._C._ImperativeEngine method
• GB0166 — Unsupported torch._C._ImperativeEngine.queue_callback()
• GB0167 — Variadic function call with bad args/kwargs type
• GB0168 — Variadic function call with bad flags
• GB0169 — Write to immutable cell
• GB0170 — Data-dependent branching
• GB0171 — assert with non-string message
• GB0172 — async_op=True for distributed collectives
• GB0173 — backward_state does not support export
• GB0174 — bad args to builtin cast()
• GB0175 — builtin isinstance() cannot determine type of argument
• GB0176 — call_id() without associated real value
• GB0177 — can’t handle functions not implemented in python
• GB0178 — constant fold exception
• GB0179 — copy.deepcopy()
• GB0180 — dataclass fields failure
• GB0181 — dtype mismatch between tensor and its gradient
• GB0182 — failed to broadcast when attempting Tensor comparison op
• GB0183 — failed to call dict.fromkeys()
• GB0184 — failed to call str() on user defined object
• GB0185 — failed to convert numpy.ndarray to Tensor
• GB0186 — functools.partial() with non-literal keyword
• GB0187 — functools.wraps
• GB0188 — getattr with no source
• GB0189 — getattr() on nn.Module with pending mutation
• GB0190 — getattr() with non-constant name argument
• GB0191 — id() with unsupported args
• GB0192 — input iterator to itertools.cycle has too many items
• GB0193 — invalid call to builtin op handler
• GB0194 — isinstance() called on user defined object with C extensions
• GB0195 — issubclass() with non-constant arguments
• GB0196 — key not found in dict
• GB0197 — list elements are pointing to the list itself
• GB0198 — mapping proxy affected by dictionary mutation
• GB0199 — mapping proxy cannot be reconstructed
• GB0200 — missing BUILD_SET handler
• GB0201 — namedtuple construction
• GB0202 — non-const argument in nn.Module method
• GB0203 — non-const keys in dict_keys
• GB0204 — non-const keys in mappingproxy
• GB0205 — proxy not set
• GB0206 — setattr() on Tensor.requires_grad
• GB0207 — sort with non-constant keys
• GB0208 — torch.* op returned non-Tensor
• GB0209 — torch.autograd._unsafe_preserve_version_counter escaped from compiled region
• GB0210 — torch.distributed package is not available!
• GB0211 — torch.nn.Module with a non-function custom getattr
• GB0212 — torch.profiler object escaped from compiled region
• GB0213 — unimplemented builtin op on tensor arguments
• GB0214 — unsupported SymNode comparison op
• GB0215 — unsupported Tensor comparison op
• GB0216 — unsupported grid type for triton hop check_grid
• GB0217 — unsupported hasattr operation
• GB0218 — unsupported index(Tensor)
• GB0219 — Backend compiler exception
• GB0220 — Failed to mutate tensor data attribute to different dtype
• GB0221 — non-generator contextlib.contextmanager
• GB0222 — Attempted to wrap a set with tensors
• GB0223 — torch.compile call with > 1 args
• GB0224 — Attempted to call torch in-graph function on only torch.SymInt arguments
• GB0225 — Attempted to use tensor creation function with requires_grad=True
• GB0226 — torch.nn.Parameter() with unsupported data type
• GB0227 — Attempted to use torch.nn.Parameter constructor with tensor subclass
• GB0228 — torch.nn.Parameter: cannot convert to traceable tracable
• GB0229 — Unexpected type of data placeholder op for parameter construction
• GB0230 — Attempted to use torch.use_deterministic_algorithms(warn_only=True)
• GB0231 — call torch.from_numpy with torch._dynamo.config.trace_numpy=False
• GB0232 — torch.from_numpy with NumPy unavailable
• GB0233 — Attempted to use strided NestedTensor
• GB0234 — Attempted to pop from empty torch function mode stack
• GB0235 — torch.nn.Parameter with non-constant Tensor attributes
• GB0236 — Invalid input type for nonstrict_trace-ed function
• GB0237 — non-constant requires_grad argument to torch.nn.Parameter
• GB0238 — Input marked with pytree.register_constant constructed in the torch.compile region
• GB0239 — Invalid use of pytree_flatten with nonstrict_trace-ed function
• GB0240 — Shape mismatch with out= list of tensor variants
• GB0241 — Attempted to call op with non-contiguous out= list of tensors
• GB0242 — Attempted to call op with non-contiguous out= tensor
• GB0243 — Attempted to use torch.nn.modules.utils._ntuple with unsupported argument type
• GB0244 — Attempted to use torch.nn.Parameter() with export
• GB0245 — Attempted to use nested_tensor with non-list input
• GB0246 — Attempted to use torch.nn.functional.one_hot with data-dependent output shape
• GB0247 — Shape mismatch with out= tensor variant
• GB0248 — improper torch.get_device_module arguments
• GB0249 — bad device argument to torch.get_device_module
• GB0250 — ndarray.astype(object)