using size() twice in a row on same tensor causes interpreter out of sync error with thunder jit

[Fuzzkatt]

Using size() twice in a row on the same tensor causes the interpreter to hit an out of sync error with thunder jit.

Minimal repro:

import torch
import thunder

def func(a):
    b = a.size()
    c = a.size() # expect error here
    return a

a = torch.randn(100, 100, device='cuda')
jfunc = thunder.jit(func)
jfunc(a)

Full error log:

Traceback (most recent call last):
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6683, in fn_
    interpretation_result: Any = _interpret_call(wrapped_fn_2, args, kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6009, in _interpret_call
    rval = _call_dispatch(compilectx, runtimectx, fn, *args, **kwargs)  # type: ignore
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6239, in _call_dispatch
    return _setup_frame_and_run_python_function(compilectx, runtimectx, wrapped_fn, *args, **kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6368, in _setup_frame_and_run_python_function
    raise e.with_traceback(tb)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6669, in fn_2
    return fn(*args, **kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6363, in _setup_frame_and_run_python_function
    res, status = _run_frame(frame, compilectx, runtimectx)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6409, in _run_frame
    interpretation_result: None | int | INTERPRETER_SIGNALS = compilectx.interpret(
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 412, in interpret
    return self._opcode_interpreter(inst, **interpreter_state)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 1231, in default_opcode_interpreter
    return handler(inst, **interpreter_state)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 3631, in _call_function_ex_handler
    return check_and_append(stack, _interpret_call(func, *args, **kwargs))
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6009, in _interpret_call
    rval = _call_dispatch(compilectx, runtimectx, fn, *args, **kwargs)  # type: ignore
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6239, in _call_dispatch
    return _setup_frame_and_run_python_function(compilectx, runtimectx, wrapped_fn, *args, **kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6368, in _setup_frame_and_run_python_function
    raise e.with_traceback(tb)
  File "/patwang-space/thunder_transform_pass/min_repro.py", line 6, in func
    c = a.size() # expect error here
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6363, in _setup_frame_and_run_python_function
    res, status = _run_frame(frame, compilectx, runtimectx)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6409, in _run_frame
    interpretation_result: None | int | INTERPRETER_SIGNALS = compilectx.interpret(
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 412, in interpret
    return self._opcode_interpreter(inst, **interpreter_state)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 1231, in default_opcode_interpreter
    return handler(inst, **interpreter_state)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 4624, in _load_method_handler
    meth = _interpret_call(getattr, obj, name)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6009, in _interpret_call
    rval = _call_dispatch(compilectx, runtimectx, fn, *args, **kwargs)  # type: ignore
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6170, in _call_dispatch
    res = lookaside_fn(*args, **kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/core/jit_ext.py", line 693, in _general_jit_getattr_lookaside
    value = getattr_lookaside(obj, name, *maybe_default)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 1684, in _getattr_lookaside
    result = wrap_attribute(result, obj, name)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 1638, in wrap_attribute
    assert plausibly_wrapper_of(
AssertionError: attribute size of TensorProxy object out of sync: <function prop_lookaside_wrap.<locals>.fn.<locals>.fn_ at 0x7f8b2ae6fac0> vs. <function prop_lookaside_wrap.<locals>.fn.<locals>.fn_ at 0x7f8b2ae6ff40>

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/patwang-space/thunder_transform_pass/min_repro.py", line 11, in <module>
    jfunc(a)
  File "/opt/pytorch/lightning-thunder/thunder/__init__.py", line 661, in fn_
    cache_entry, inps, pro_to_epi = get_computation_and_inputs(*args, **kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/__init__.py", line 277, in cache_info_wrapper
    res = fn(*args, **kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/__init__.py", line 538, in get_computation_and_inputs
    jit_results: TraceResults = interpreter(
  File "/opt/pytorch/lightning-thunder/thunder/__init__.py", line 190, in _general_frontend
    return thunder_general_jit(fn, args, kwargs, sharp_edges=sharp_edges, record_history=record_history)
  File "/opt/pytorch/lightning-thunder/thunder/core/jit_ext.py", line 1510, in thunder_general_jit
    result = jfn(*args, **kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/core/interpreter.py", line 6692, in fn_
    raise InterpreterError(msg) from e
thunder.core.interpreter.InterpreterError: Encountered exception AssertionError: attribute size of TensorProxy object out of sync: <function prop_lookaside_wrap.<locals>.fn.<locals>.fn_ at 0x7f8b2ae6fac0> vs. <function prop_lookaside_wrap.<locals>.fn.<locals>.fn_ at 0x7f8b2ae6ff40> while tracing <function func at 0x7f8bf9163d90>:

[Fuzzkatt]

Note that calling size twice in the same trace but on different tensors works fine:

import torch
import thunder

def func(a, b):
    c = a.size()
    d = b.size() # works
    return a, b

a = torch.randn(100, 100, device='cuda')
b = torch.randn(100, 100, device='cuda')
jfunc = thunder.jit(func)
jfunc(a, b)

Additionally, calling size once in the same trace but invoking the trace twice is also fine:

import torch
import thunder

def func(a):
    b = a.size()
    return a

a = torch.randn(100, 100, device='cuda')
jfunc = thunder.jit(func)
jfunc(a)
jfunc(a) # works