評價此頁
★ ★ ★ ★ ★

常見圖中斷#

建立時間：2025 年 7 月 28 日 | 最後更新時間：2025 年 7 月 28 日

以下是一些常見的圖中斷及其解決方法。

錯誤程式碼#

您的程式碼可能包含錯誤（意味著即使沒有 torch.compile 也會導致錯誤）。在下面的示例中，由於多餘的引數，torch.sin 呼叫中存在拼寫錯誤。始終停用 torch.compile 以檢查程式碼是否能正確執行。

@torch.compile
def fn(x):
    y = torch.sin(x, x)
    return y

try:
    fn(torch.ones(3, 3))
except Exception as e:
    pass

Graph break in user code at /tmp/ipykernel_581/343837593.py:3
Graph Break Reason: TypeError when making fake tensor call
  Explanation: 


  Developer debug context: TypeError <built-in method sin of type object at 0x7f08e7582260>: sin() takes 1 positional argument but 2 were given

 For more details about this graph break, please visit: https://meta-pytorch.github.io/compile-graph-break-site/gb/gb0112.html
User code traceback:
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel_launcher.py", line 18, in <module>
    app.launch_new_instance()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/traitlets/config/application.py", line 1075, in launch_instance
    app.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelapp.py", line 739, in start
    self.io_loop.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/tornado/platform/asyncio.py", line 211, in start
    self.asyncio_loop.run_forever()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 603, in run_forever
    self._run_once()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 1909, in _run_once
    handle._run()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/events.py", line 80, in _run
    self._context.run(self._callback, *self._args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 519, in dispatch_queue
    await self.process_one()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 508, in process_one
    await dispatch(*args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 400, in dispatch_shell
    await result
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 368, in execute_request
    await super().execute_request(stream, ident, parent)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 767, in execute_request
    reply_content = await reply_content
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 455, in do_execute
    res = shell.run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/zmqshell.py", line 577, in run_cell
    return super().run_cell(*args, **kwargs)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3006, in run_cell
    result = self._run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3061, in _run_cell
    result = runner(coro)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/async_helpers.py", line 129, in _pseudo_sync_runner
    coro.send(None)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3266, in run_cell_async
    has_raised = await self.run_ast_nodes(code_ast.body, cell_name,
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3445, in run_ast_nodes
    if await self.run_code(code, result, async_=asy):
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3505, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "/tmp/ipykernel_581/343837593.py", line 7, in <module>
    fn(torch.ones(3, 3))
  File "/tmp/ipykernel_581/343837593.py", line 3, in fn
    y = torch.sin(x, x)

Dynamo 會盡力提示圖中斷是否由您的程式碼引起。但有時仍然很難從日誌中判斷圖中斷是由程式碼錯誤、更復雜的圖中斷還是 torch.compile bug 引起的。為了區分，我們建議嘗試在不使用 torch.compile 的情況下執行您的程式碼，看看是否仍然會收到圖中斷報告的錯誤。

依賴資料的操作#

torch.compile 會在依賴資料的操作上發生圖中斷，例如依賴資料的控制流（if 語句、帶有張量的迴圈）以及直接訪問張量資料（.item, .data_ptr）。

@torch.compile
def fn(x):
    y = x.sum()
    if y > 0:
        return x + y.item()
    return x - y.item()

print(fn(torch.ones(3, 3)))

tensor([[10., 10., 10.],
        [10., 10., 10.],
        [10., 10., 10.]])

Graph break in user code at /tmp/ipykernel_581/3495555842.py:4
Graph Break Reason: Data-dependent branching
  Explanation: Detected data-dependent branching (e.g. `if my_tensor.sum() > 0:`). Dynamo does not support tracing dynamic control flow.
  Hint: This graph break is fundamental - it is unlikely that Dynamo will ever be able to trace through your code. Consider finding a workaround.
  Hint: Use `torch.cond` to express dynamic control flow.

  Developer debug context: attempted to jump with TensorVariable()

User code traceback:
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel_launcher.py", line 18, in <module>
    app.launch_new_instance()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/traitlets/config/application.py", line 1075, in launch_instance
    app.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelapp.py", line 739, in start
    self.io_loop.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/tornado/platform/asyncio.py", line 211, in start
    self.asyncio_loop.run_forever()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 603, in run_forever
    self._run_once()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 1909, in _run_once
    handle._run()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/events.py", line 80, in _run
    self._context.run(self._callback, *self._args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 519, in dispatch_queue
    await self.process_one()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 508, in process_one
    await dispatch(*args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 400, in dispatch_shell
    await result
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 368, in execute_request
    await super().execute_request(stream, ident, parent)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 767, in execute_request
    reply_content = await reply_content
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 455, in do_execute
    res = shell.run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/zmqshell.py", line 577, in run_cell
    return super().run_cell(*args, **kwargs)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3006, in run_cell
    result = self._run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3061, in _run_cell
    result = runner(coro)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/async_helpers.py", line 129, in _pseudo_sync_runner
    coro.send(None)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3266, in run_cell_async
    has_raised = await self.run_ast_nodes(code_ast.body, cell_name,
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3445, in run_ast_nodes
    if await self.run_code(code, result, async_=asy):
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3505, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "/tmp/ipykernel_581/3495555842.py", line 8, in <module>
    print(fn(torch.ones(3, 3)))
  File "/tmp/ipykernel_581/3495555842.py", line 4, in fn
    if y > 0:

Graph break from `Tensor.item()`, consider setting:
    torch._dynamo.config.capture_scalar_outputs = True
or:
    env TORCHDYNAMO_CAPTURE_SCALAR_OUTPUTS=1
to include these operations in the captured graph.

Graph break: from user code at:
  File "/tmp/ipykernel_581/3495555842.py", line 5, in torch_dynamo_resume_in_fn_at_4
    return x + y.item()


Graph break in user code at /tmp/ipykernel_581/3495555842.py:5
Graph Break Reason: Unsupported Tensor.item() call with capture_scalar_outputs=False
  Explanation: Dynamo does not support tracing `Tensor.item()` with config.capture_scalar_outputs=False.
  Hint: Set `torch._dynamo.config.capture_scalar_outputs = True` or `export TORCHDYNAMO_CAPTURE_SCALAR_OUTPUTS=1` to include these operations in the captured graph.

  Developer debug context: call_method TensorVariable() item () {}

 For more details about this graph break, please visit: https://meta-pytorch.github.io/compile-graph-break-site/gb/gb0124.html
User code traceback:
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel_launcher.py", line 18, in <module>
    app.launch_new_instance()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/traitlets/config/application.py", line 1075, in launch_instance
    app.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelapp.py", line 739, in start
    self.io_loop.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/tornado/platform/asyncio.py", line 211, in start
    self.asyncio_loop.run_forever()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 603, in run_forever
    self._run_once()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 1909, in _run_once
    handle._run()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/events.py", line 80, in _run
    self._context.run(self._callback, *self._args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 519, in dispatch_queue
    await self.process_one()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 508, in process_one
    await dispatch(*args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 400, in dispatch_shell
    await result
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 368, in execute_request
    await super().execute_request(stream, ident, parent)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 767, in execute_request
    reply_content = await reply_content
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 455, in do_execute
    res = shell.run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/zmqshell.py", line 577, in run_cell
    return super().run_cell(*args, **kwargs)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3006, in run_cell
    result = self._run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3061, in _run_cell
    result = runner(coro)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/async_helpers.py", line 129, in _pseudo_sync_runner
    coro.send(None)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3266, in run_cell_async
    has_raised = await self.run_ast_nodes(code_ast.body, cell_name,
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3445, in run_ast_nodes
    if await self.run_code(code, result, async_=asy):
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3505, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "/tmp/ipykernel_581/3495555842.py", line 8, in <module>
    print(fn(torch.ones(3, 3)))
  File "/tmp/ipykernel_581/3495555842.py", line 5, in fn
    return x + y.item()

這些圖中斷的通用解決方法是避免進行依賴資料的操作。一些具體的解決方法是：

  • 如果您的控制流實際上不依賴於資料值，請考慮修改程式碼以對常量執行控制流。

# old
x = torch.randn(3, 3)
@torch.compile
def fn(y):
    if x.sum() > 0:
        return y + x
    else:
        return y - x

print(fn(torch.ones(3, 3)))

tensor([[3.6763, 1.8030, 1.3392],
        [1.4292, 0.4356, 2.3070],
        [1.9438, 1.3472, 0.3239]])

Graph break in user code at /tmp/ipykernel_581/2410325100.py:5
Graph Break Reason: Data-dependent branching
  Explanation: Detected data-dependent branching (e.g. `if my_tensor.sum() > 0:`). Dynamo does not support tracing dynamic control flow.
  Hint: This graph break is fundamental - it is unlikely that Dynamo will ever be able to trace through your code. Consider finding a workaround.
  Hint: Use `torch.cond` to express dynamic control flow.

  Developer debug context: attempted to jump with TensorVariable()

User code traceback:
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel_launcher.py", line 18, in <module>
    app.launch_new_instance()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/traitlets/config/application.py", line 1075, in launch_instance
    app.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelapp.py", line 739, in start
    self.io_loop.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/tornado/platform/asyncio.py", line 211, in start
    self.asyncio_loop.run_forever()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 603, in run_forever
    self._run_once()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 1909, in _run_once
    handle._run()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/events.py", line 80, in _run
    self._context.run(self._callback, *self._args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 519, in dispatch_queue
    await self.process_one()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 508, in process_one
    await dispatch(*args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 400, in dispatch_shell
    await result
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 368, in execute_request
    await super().execute_request(stream, ident, parent)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 767, in execute_request
    reply_content = await reply_content
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 455, in do_execute
    res = shell.run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/zmqshell.py", line 577, in run_cell
    return super().run_cell(*args, **kwargs)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3006, in run_cell
    result = self._run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3061, in _run_cell
    result = runner(coro)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/async_helpers.py", line 129, in _pseudo_sync_runner
    coro.send(None)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3266, in run_cell_async
    has_raised = await self.run_ast_nodes(code_ast.body, cell_name,
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3445, in run_ast_nodes
    if await self.run_code(code, result, async_=asy):
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3505, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "/tmp/ipykernel_581/2410325100.py", line 10, in <module>
    print(fn(torch.ones(3, 3)))
  File "/tmp/ipykernel_581/2410325100.py", line 5, in fn
    if x.sum() > 0:

# new
x = torch.randn(3, 3)
cond = (x.sum() > 0).item()
@torch.compile
def fn(y):
    if cond:
        return y + x
    else:
        return y - x

print(fn(torch.ones(3, 3)))

tensor([[2.8814, 1.2171, 1.6990],
        [1.0319, 2.2353, 0.8251],
        [0.0565, 0.7153, 0.7048]])

  • 使用更高級別的操作，例如 控制流 - Cond，來代替依賴資料的控制流。

# old
@torch.compile
def fn(x):
    if x.sum() > 0:
        return x + 1
    return x - 1

print(fn(torch.ones(3, 3)))

tensor([[2., 2., 2.],
        [2., 2., 2.],
        [2., 2., 2.]])

Graph break in user code at /tmp/ipykernel_581/520574912.py:4
Graph Break Reason: Data-dependent branching
  Explanation: Detected data-dependent branching (e.g. `if my_tensor.sum() > 0:`). Dynamo does not support tracing dynamic control flow.
  Hint: This graph break is fundamental - it is unlikely that Dynamo will ever be able to trace through your code. Consider finding a workaround.
  Hint: Use `torch.cond` to express dynamic control flow.

  Developer debug context: attempted to jump with TensorVariable()

User code traceback:
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/opt/conda/envs/py_3.10/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel_launcher.py", line 18, in <module>
    app.launch_new_instance()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/traitlets/config/application.py", line 1075, in launch_instance
    app.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelapp.py", line 739, in start
    self.io_loop.start()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/tornado/platform/asyncio.py", line 211, in start
    self.asyncio_loop.run_forever()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 603, in run_forever
    self._run_once()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/base_events.py", line 1909, in _run_once
    handle._run()
  File "/opt/conda/envs/py_3.10/lib/python3.10/asyncio/events.py", line 80, in _run
    self._context.run(self._callback, *self._args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 519, in dispatch_queue
    await self.process_one()
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 508, in process_one
    await dispatch(*args)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 400, in dispatch_shell
    await result
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 368, in execute_request
    await super().execute_request(stream, ident, parent)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/kernelbase.py", line 767, in execute_request
    reply_content = await reply_content
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/ipkernel.py", line 455, in do_execute
    res = shell.run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/ipykernel/zmqshell.py", line 577, in run_cell
    return super().run_cell(*args, **kwargs)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3006, in run_cell
    result = self._run_cell(
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3061, in _run_cell
    result = runner(coro)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/async_helpers.py", line 129, in _pseudo_sync_runner
    coro.send(None)
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3266, in run_cell_async
    has_raised = await self.run_ast_nodes(code_ast.body, cell_name,
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3445, in run_ast_nodes
    if await self.run_code(code, result, async_=asy):
  File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3505, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "/tmp/ipykernel_581/520574912.py", line 8, in <module>
    print(fn(torch.ones(3, 3)))
  File "/tmp/ipykernel_581/520574912.py", line 4, in fn
    if x.sum() > 0:

# new
@torch.compile
def fn(x):
    return torch.cond(
        x.sum() > 0,
        lambda x: x + 1,
        lambda x: x - 1,
        (x,),
    )

print(fn(torch.ones(3, 3)))

tensor([[2., 2., 2.],
        [2., 2., 2.],
        [2., 2., 2.]])

  • 如果您有一個 .item() 呼叫，可以嘗試使用 torch._dynamo.config.capture_scalar_outputs = True 或 TORCHDYNAMO_CAPTURE_SCALAR_OUTPUTS=1。

  • 將函式中有問題的部分包裝成自定義運算子。

列印和日誌記錄#

列印/記錄/發出警告將導致圖中斷。您可以嘗試使用 torch._dynamo.config.reorderable_logging_functions 來解決這個問題。此配置用於重新排序日誌記錄函式，使它們在跟蹤函式的末尾被呼叫，從而避免圖中斷。但是，如果發生變異等情況，記錄的內容可能會有所不同。

torch._dynamo.config.reorderable_logging_functions.add(print)

@torch.compile
def fn(x):
    x += 1
    print("log!")
    return torch.sin(x)

print(fn(torch.ones(3, 3)))

log!
tensor([[0.9093, 0.9093, 0.9093],
        [0.9093, 0.9093, 0.9093],
        [0.9093, 0.9093, 0.9093]])
在此頁面上
在 GitHub 上編輯
顯示原始碼
PyTorch 庫