Mechanisms to register components of the code base that have debugging functionality #3549
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…debugging features that will be active in a context manager (compile time only for now)
github-actions
bot
added
component: lowering
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
There are some changes that do not conform to Python style guidelines:
--- /home/runner/work/TensorRT/TensorRT/examples/dynamo/debugging_example.py 2025-06-03 04:51:19.066284+00:00
+++ /home/runner/work/TensorRT/TensorRT/examples/dynamo/debugging_example.py 2025-06-03 04:51:41.306273+00:00
@@ -27,14 +27,16 @@
# %%
# Compilation with `torch_tensorrt.compile`
# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
# Compile within the debugging context to control different aspects of the compilation pipeline
-with torch_tensorrt.dynamo._Debugger.Debugger(engine_builder_monitor=True, break_in_remove_assert_nodes=True):
+with torch_tensorrt.dynamo._Debugger.Debugger(
+ engine_builder_monitor=True, break_in_remove_assert_nodes=True
+):
optimized_model = torch_tensorrt.compile(
model,
- #ir="torch_compile",
+ # ir="torch_compile",
inputs=inputs,
enabled_precisions=enabled_precisions,
workspace_size=workspace_size,
min_block_size=min_block_size,
)
@@ -50,11 +52,11 @@
print(new_outputs)
optimized_model = torch_tensorrt.compile(
model,
- #ir="torch_compile",
+ # ir="torch_compile",
inputs=inputs,
enabled_precisions=enabled_precisions,
workspace_size=workspace_size,
min_block_size=min_block_size,
)
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_DebuggerConfig.py 2025-06-03 04:51:19.076284+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_DebuggerConfig.py 2025-06-03 04:51:41.757086+00:00
@@ -1,13 +1,14 @@
from typing import Any, List, Optional, Dict
import logging
from dataclasses import dataclass, field
+
@dataclass
class DebuggerConfig:
- log_level: int = logging.getLevelName('DEBUG')
+ log_level: int = logging.getLevelName("DEBUG")
capture_fx_graph_before: List[str] = field(default_factory=lambda: [])
capture_fx_graph_after: List[str] = field(default_factory=lambda: [])
save_engine_profile: bool = False
engine_builder_monitor: bool = True
break_in_remove_assert_nodes: bool = True
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_Debugger.py 2025-06-03 04:51:19.076284+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_Debugger.py 2025-06-03 04:51:41.803440+00:00
@@ -11,11 +11,14 @@
from logging.config import dictConfig
import torch
import torch_tensorrt
from torch_tensorrt.dynamo._DebuggerConfig import DebuggerConfig
-from torch_tensorrt.dynamo._supports_debugger import _DEBUG_ENABLED_CLS, _DEBUG_ENABLED_FUNCS
+from torch_tensorrt.dynamo._supports_debugger import (
+ _DEBUG_ENABLED_CLS,
+ _DEBUG_ENABLED_FUNCS,
+)
from torch_tensorrt.dynamo.conversion._TRTInterpreter import TRTInterpreter
from torch_tensorrt.dynamo.lowering.passes.constant_folding import constant_fold
from torch_tensorrt.dynamo.lowering import (
ATEN_POST_LOWERING_PASSES,
ATEN_PRE_LOWERING_PASSES,
@@ -33,46 +36,52 @@
self.cfg = DebuggerConfig(**kwargs)
def __enter__(self) -> None:
self.original_lvl = _LOGGER.getEffectiveLevel()
self.rt_level = torch.ops.tensorrt.get_logging_level()
- #dictConfig(self.get_config())
+ # dictConfig(self.get_config())
self.old_pre_passes, self.old_post_passes = (
ATEN_PRE_LOWERING_PASSES.passes,
ATEN_POST_LOWERING_PASSES.passes,
)
pre_pass_names = [p.__name__ for p in self.old_pre_passes]
post_pass_names = [p.__name__ for p in self.old_post_passes]
- #path = os.path.join(DEBUG_FILE_DIR, "lowering_passes_visualization")
+ # path = os.path.join(DEBUG_FILE_DIR, "lowering_passes_visualization")
- ATEN_PRE_LOWERING_PASSES.insert_debug_pass_before(self.cfg.capture_fx_graph_before)
- ATEN_POST_LOWERING_PASSES.insert_debug_pass_before(self.cfg.capture_fx_graph_before)
+ ATEN_PRE_LOWERING_PASSES.insert_debug_pass_before(
+ self.cfg.capture_fx_graph_before
+ )
+ ATEN_POST_LOWERING_PASSES.insert_debug_pass_before(
+ self.cfg.capture_fx_graph_before
+ )
- ATEN_PRE_LOWERING_PASSES.insert_debug_pass_after(self.cfg.capture_fx_graph_after)
- ATEN_POST_LOWERING_PASSES.insert_debug_pass_after(self.cfg.capture_fx_graph_after)
+ ATEN_PRE_LOWERING_PASSES.insert_debug_pass_after(
+ self.cfg.capture_fx_graph_after
+ )
+ ATEN_POST_LOWERING_PASSES.insert_debug_pass_after(
+ self.cfg.capture_fx_graph_after
+ )
self._context_stack = contextlib.ExitStack()
for f in _DEBUG_ENABLED_FUNCS:
f.__kwdefaults__["_debugger_settings"] = self.cfg
[
self._context_stack.enter_context(
mock.patch.object(
c,
- '__init__',
- functools.partialmethod(
- c.__init__,
- _debugger_settings=self.cfg
- )
+ "__init__",
+ functools.partialmethod(c.__init__, _debugger_settings=self.cfg),
)
- ) for c in _DEBUG_ENABLED_CLS
+ )
+ for c in _DEBUG_ENABLED_CLS
]
def __exit__(self, exc_type: Any, exc_value: Any, exc_tb: Any) -> None:
- #dictConfig(self.get_default_config())
+ # dictConfig(self.get_default_config())
torch.ops.tensorrt.set_logging_level(self.rt_level)
ATEN_PRE_LOWERING_PASSES.passes, ATEN_POST_LOWERING_PASSES.passes = (
self.old_pre_passes,
self.old_post_passes,
@@ -80,12 +89,10 @@
for f in _DEBUG_ENABLED_FUNCS:
f.__kwdefaults__["_debugger_settings"] = None
self._context_stack.close()
-
-
# def get_config(self) -> dict[str, Any]:
# config = {
# "version": 1,
# "disable_existing_loggers": False,
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_supports_debugger.py 2025-06-03 04:51:19.077284+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_supports_debugger.py 2025-06-03 04:51:41.917136+00:00
@@ -4,12 +4,14 @@
_DEBUG_ENABLED_FUNCS = []
_DEBUG_ENABLED_CLS = []
+
def fn_supports_debugger(func):
_DEBUG_ENABLED_FUNCS.append(func)
return func
-def cls_supports_debugger(cls: Type[T]) -> Type[T]:
+
+def cls_supports_debugger(cls: Type[T]) -> Type[T]:
_DEBUG_ENABLED_CLS.append(cls)
return cls
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTBuilderMonitor.py 2025-06-03 04:51:19.077284+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTBuilderMonitor.py 2025-06-03 04:51:42.103174+00:00
@@ -90,10 +90,11 @@
clear_line()
print()
move_cursor_up(len(self._active_phases) + blank_lines)
sys.stdout.flush()
+
# try:
# from rich.progress import BarColumn, Progress, TaskID, TextColumn, TimeElapsedColumn
# class _RichMonitor(trt.IProgressMonitor): # type: ignore
# def __init__(self, engine_name: str = "") -> None:
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py 2025-06-03 04:51:19.078285+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py 2025-06-03 04:51:42.829249+00:00
@@ -69,10 +69,11 @@
input_names: Sequence[str]
output_names: Sequence[str]
weight_name_map: Optional[dict[Any, Any]]
requires_output_allocator: bool
+
@cls_supports_debugger
class TRTInterpreter(torch.fx.Interpreter): # type: ignore[misc]
def __init__(
self,
module: torch.fx.GraphModule,
@@ -207,11 +208,14 @@
algorithm_selector: Optional[trt.IAlgorithmSelector] = None,
tactic_sources: Optional[int] = None,
) -> trt.IBuilderConfig:
builder_config = self.builder.create_builder_config()
- if self._debugger_settings is not None and self._debugger_settings.engine_builder_monitor:
+ if (
+ self._debugger_settings is not None
+ and self._debugger_settings.engine_builder_monitor
+ ):
builder_config.progress_monitor = TRTBulderMonitor()
if self.compilation_settings.workspace_size != 0:
builder_config.set_memory_pool_limit(
trt.MemoryPoolType.WORKSPACE, self.compilation_settings.workspace_size
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/remove_assert_nodes.py 2025-06-03 04:51:19.081285+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/remove_assert_nodes.py 2025-06-03 04:51:43.303996+00:00
@@ -9,16 +9,23 @@
clean_up_graph_after_modifications,
)
logger = logging.getLogger(__name__)
+
@fn_supports_debugger
def remove_assert_nodes(
- gm: torch.fx.GraphModule, settings: CompilationSettings, *, _debugger_settings: Optional[DebuggerConfig]=None
+ gm: torch.fx.GraphModule,
+ settings: CompilationSettings,
+ *,
+ _debugger_settings: Optional[DebuggerConfig] = None,
) -> torch.fx.GraphModule:
"""Remove assert_scalar ops in the graph"""
- if _debugger_settings is not None and _debugger_settings.break_in_remove_assert_nodes:
+ if (
+ _debugger_settings is not None
+ and _debugger_settings.break_in_remove_assert_nodes
+ ):
breakpoint()
count = 0
for node in gm.graph.nodes:
if (
node.target == torch.ops.aten._assert_scalar.default
There was a problem hiding this comment.
There are some changes that do not conform to Python style guidelines:
--- /home/runner/work/TensorRT/TensorRT/examples/dynamo/debugging_example.py 2025-06-03 04:51:19.389777+00:00
+++ /home/runner/work/TensorRT/TensorRT/examples/dynamo/debugging_example.py 2025-06-03 04:51:41.613814+00:00
@@ -27,14 +27,16 @@
# %%
# Compilation with `torch_tensorrt.compile`
# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
# Compile within the debugging context to control different aspects of the compilation pipeline
-with torch_tensorrt.dynamo._Debugger.Debugger(engine_builder_monitor=True, break_in_remove_assert_nodes=True):
+with torch_tensorrt.dynamo._Debugger.Debugger(
+ engine_builder_monitor=True, break_in_remove_assert_nodes=True
+):
optimized_model = torch_tensorrt.compile(
model,
- #ir="torch_compile",
+ # ir="torch_compile",
inputs=inputs,
enabled_precisions=enabled_precisions,
workspace_size=workspace_size,
min_block_size=min_block_size,
)
@@ -50,11 +52,11 @@
print(new_outputs)
optimized_model = torch_tensorrt.compile(
model,
- #ir="torch_compile",
+ # ir="torch_compile",
inputs=inputs,
enabled_precisions=enabled_precisions,
workspace_size=workspace_size,
min_block_size=min_block_size,
)
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_DebuggerConfig.py 2025-06-03 04:51:19.400778+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_DebuggerConfig.py 2025-06-03 04:51:42.057169+00:00
@@ -1,13 +1,14 @@
from typing import Any, List, Optional, Dict
import logging
from dataclasses import dataclass, field
+
@dataclass
class DebuggerConfig:
- log_level: int = logging.getLevelName('DEBUG')
+ log_level: int = logging.getLevelName("DEBUG")
capture_fx_graph_before: List[str] = field(default_factory=lambda: [])
capture_fx_graph_after: List[str] = field(default_factory=lambda: [])
save_engine_profile: bool = False
engine_builder_monitor: bool = True
break_in_remove_assert_nodes: bool = True
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_Debugger.py 2025-06-03 04:51:19.400778+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_Debugger.py 2025-06-03 04:51:42.084847+00:00
@@ -11,11 +11,14 @@
from logging.config import dictConfig
import torch
import torch_tensorrt
from torch_tensorrt.dynamo._DebuggerConfig import DebuggerConfig
-from torch_tensorrt.dynamo._supports_debugger import _DEBUG_ENABLED_CLS, _DEBUG_ENABLED_FUNCS
+from torch_tensorrt.dynamo._supports_debugger import (
+ _DEBUG_ENABLED_CLS,
+ _DEBUG_ENABLED_FUNCS,
+)
from torch_tensorrt.dynamo.conversion._TRTInterpreter import TRTInterpreter
from torch_tensorrt.dynamo.lowering.passes.constant_folding import constant_fold
from torch_tensorrt.dynamo.lowering import (
ATEN_POST_LOWERING_PASSES,
ATEN_PRE_LOWERING_PASSES,
@@ -33,46 +36,52 @@
self.cfg = DebuggerConfig(**kwargs)
def __enter__(self) -> None:
self.original_lvl = _LOGGER.getEffectiveLevel()
self.rt_level = torch.ops.tensorrt.get_logging_level()
- #dictConfig(self.get_config())
+ # dictConfig(self.get_config())
self.old_pre_passes, self.old_post_passes = (
ATEN_PRE_LOWERING_PASSES.passes,
ATEN_POST_LOWERING_PASSES.passes,
)
pre_pass_names = [p.__name__ for p in self.old_pre_passes]
post_pass_names = [p.__name__ for p in self.old_post_passes]
- #path = os.path.join(DEBUG_FILE_DIR, "lowering_passes_visualization")
+ # path = os.path.join(DEBUG_FILE_DIR, "lowering_passes_visualization")
- ATEN_PRE_LOWERING_PASSES.insert_debug_pass_before(self.cfg.capture_fx_graph_before)
- ATEN_POST_LOWERING_PASSES.insert_debug_pass_before(self.cfg.capture_fx_graph_before)
+ ATEN_PRE_LOWERING_PASSES.insert_debug_pass_before(
+ self.cfg.capture_fx_graph_before
+ )
+ ATEN_POST_LOWERING_PASSES.insert_debug_pass_before(
+ self.cfg.capture_fx_graph_before
+ )
- ATEN_PRE_LOWERING_PASSES.insert_debug_pass_after(self.cfg.capture_fx_graph_after)
- ATEN_POST_LOWERING_PASSES.insert_debug_pass_after(self.cfg.capture_fx_graph_after)
+ ATEN_PRE_LOWERING_PASSES.insert_debug_pass_after(
+ self.cfg.capture_fx_graph_after
+ )
+ ATEN_POST_LOWERING_PASSES.insert_debug_pass_after(
+ self.cfg.capture_fx_graph_after
+ )
self._context_stack = contextlib.ExitStack()
for f in _DEBUG_ENABLED_FUNCS:
f.__kwdefaults__["_debugger_settings"] = self.cfg
[
self._context_stack.enter_context(
mock.patch.object(
c,
- '__init__',
- functools.partialmethod(
- c.__init__,
- _debugger_settings=self.cfg
- )
+ "__init__",
+ functools.partialmethod(c.__init__, _debugger_settings=self.cfg),
)
- ) for c in _DEBUG_ENABLED_CLS
+ )
+ for c in _DEBUG_ENABLED_CLS
]
def __exit__(self, exc_type: Any, exc_value: Any, exc_tb: Any) -> None:
- #dictConfig(self.get_default_config())
+ # dictConfig(self.get_default_config())
torch.ops.tensorrt.set_logging_level(self.rt_level)
ATEN_PRE_LOWERING_PASSES.passes, ATEN_POST_LOWERING_PASSES.passes = (
self.old_pre_passes,
self.old_post_passes,
@@ -80,12 +89,10 @@
for f in _DEBUG_ENABLED_FUNCS:
f.__kwdefaults__["_debugger_settings"] = None
self._context_stack.close()
-
-
# def get_config(self) -> dict[str, Any]:
# config = {
# "version": 1,
# "disable_existing_loggers": False,
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_supports_debugger.py 2025-06-03 04:51:19.400778+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_supports_debugger.py 2025-06-03 04:51:42.196192+00:00
@@ -4,12 +4,14 @@
_DEBUG_ENABLED_FUNCS = []
_DEBUG_ENABLED_CLS = []
+
def fn_supports_debugger(func):
_DEBUG_ENABLED_FUNCS.append(func)
return func
-def cls_supports_debugger(cls: Type[T]) -> Type[T]:
+
+def cls_supports_debugger(cls: Type[T]) -> Type[T]:
_DEBUG_ENABLED_CLS.append(cls)
return cls
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTBuilderMonitor.py 2025-06-03 04:51:19.401778+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTBuilderMonitor.py 2025-06-03 04:51:42.406536+00:00
@@ -90,10 +90,11 @@
clear_line()
print()
move_cursor_up(len(self._active_phases) + blank_lines)
sys.stdout.flush()
+
# try:
# from rich.progress import BarColumn, Progress, TaskID, TextColumn, TimeElapsedColumn
# class _RichMonitor(trt.IProgressMonitor): # type: ignore
# def __init__(self, engine_name: str = "") -> None:
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py 2025-06-03 04:51:19.401778+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py 2025-06-03 04:51:43.112812+00:00
@@ -69,10 +69,11 @@
input_names: Sequence[str]
output_names: Sequence[str]
weight_name_map: Optional[dict[Any, Any]]
requires_output_allocator: bool
+
@cls_supports_debugger
class TRTInterpreter(torch.fx.Interpreter): # type: ignore[misc]
def __init__(
self,
module: torch.fx.GraphModule,
@@ -207,11 +208,14 @@
algorithm_selector: Optional[trt.IAlgorithmSelector] = None,
tactic_sources: Optional[int] = None,
) -> trt.IBuilderConfig:
builder_config = self.builder.create_builder_config()
- if self._debugger_settings is not None and self._debugger_settings.engine_builder_monitor:
+ if (
+ self._debugger_settings is not None
+ and self._debugger_settings.engine_builder_monitor
+ ):
builder_config.progress_monitor = TRTBulderMonitor()
if self.compilation_settings.workspace_size != 0:
builder_config.set_memory_pool_limit(
trt.MemoryPoolType.WORKSPACE, self.compilation_settings.workspace_size
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/remove_assert_nodes.py 2025-06-03 04:51:19.405777+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/remove_assert_nodes.py 2025-06-03 04:51:43.491038+00:00
@@ -9,16 +9,23 @@
clean_up_graph_after_modifications,
)
logger = logging.getLogger(__name__)
+
@fn_supports_debugger
def remove_assert_nodes(
- gm: torch.fx.GraphModule, settings: CompilationSettings, *, _debugger_settings: Optional[DebuggerConfig]=None
+ gm: torch.fx.GraphModule,
+ settings: CompilationSettings,
+ *,
+ _debugger_settings: Optional[DebuggerConfig] = None,
) -> torch.fx.GraphModule:
"""Remove assert_scalar ops in the graph"""
- if _debugger_settings is not None and _debugger_settings.break_in_remove_assert_nodes:
+ if (
+ _debugger_settings is not None
+ and _debugger_settings.break_in_remove_assert_nodes
+ ):
breakpoint()
count = 0
for node in gm.graph.nodes:
if (
node.target == torch.ops.aten._assert_scalar.default
There was a problem hiding this comment.
There are some changes that do not conform to Python style guidelines:
--- /home/runner/work/TensorRT/TensorRT/examples/dynamo/debugging_example.py 2025-06-03 04:51:20.312103+00:00
+++ /home/runner/work/TensorRT/TensorRT/examples/dynamo/debugging_example.py 2025-06-03 04:51:42.612724+00:00
@@ -27,14 +27,16 @@
# %%
# Compilation with `torch_tensorrt.compile`
# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
# Compile within the debugging context to control different aspects of the compilation pipeline
-with torch_tensorrt.dynamo._Debugger.Debugger(engine_builder_monitor=True, break_in_remove_assert_nodes=True):
+with torch_tensorrt.dynamo._Debugger.Debugger(
+ engine_builder_monitor=True, break_in_remove_assert_nodes=True
+):
optimized_model = torch_tensorrt.compile(
model,
- #ir="torch_compile",
+ # ir="torch_compile",
inputs=inputs,
enabled_precisions=enabled_precisions,
workspace_size=workspace_size,
min_block_size=min_block_size,
)
@@ -50,11 +52,11 @@
print(new_outputs)
optimized_model = torch_tensorrt.compile(
model,
- #ir="torch_compile",
+ # ir="torch_compile",
inputs=inputs,
enabled_precisions=enabled_precisions,
workspace_size=workspace_size,
min_block_size=min_block_size,
)
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_DebuggerConfig.py 2025-06-03 04:51:20.323103+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_DebuggerConfig.py 2025-06-03 04:51:43.069284+00:00
@@ -1,13 +1,14 @@
from typing import Any, List, Optional, Dict
import logging
from dataclasses import dataclass, field
+
@dataclass
class DebuggerConfig:
- log_level: int = logging.getLevelName('DEBUG')
+ log_level: int = logging.getLevelName("DEBUG")
capture_fx_graph_before: List[str] = field(default_factory=lambda: [])
capture_fx_graph_after: List[str] = field(default_factory=lambda: [])
save_engine_profile: bool = False
engine_builder_monitor: bool = True
break_in_remove_assert_nodes: bool = True
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_Debugger.py 2025-06-03 04:51:20.323103+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_Debugger.py 2025-06-03 04:51:43.098669+00:00
@@ -11,11 +11,14 @@
from logging.config import dictConfig
import torch
import torch_tensorrt
from torch_tensorrt.dynamo._DebuggerConfig import DebuggerConfig
-from torch_tensorrt.dynamo._supports_debugger import _DEBUG_ENABLED_CLS, _DEBUG_ENABLED_FUNCS
+from torch_tensorrt.dynamo._supports_debugger import (
+ _DEBUG_ENABLED_CLS,
+ _DEBUG_ENABLED_FUNCS,
+)
from torch_tensorrt.dynamo.conversion._TRTInterpreter import TRTInterpreter
from torch_tensorrt.dynamo.lowering.passes.constant_folding import constant_fold
from torch_tensorrt.dynamo.lowering import (
ATEN_POST_LOWERING_PASSES,
ATEN_PRE_LOWERING_PASSES,
@@ -33,46 +36,52 @@
self.cfg = DebuggerConfig(**kwargs)
def __enter__(self) -> None:
self.original_lvl = _LOGGER.getEffectiveLevel()
self.rt_level = torch.ops.tensorrt.get_logging_level()
- #dictConfig(self.get_config())
+ # dictConfig(self.get_config())
self.old_pre_passes, self.old_post_passes = (
ATEN_PRE_LOWERING_PASSES.passes,
ATEN_POST_LOWERING_PASSES.passes,
)
pre_pass_names = [p.__name__ for p in self.old_pre_passes]
post_pass_names = [p.__name__ for p in self.old_post_passes]
- #path = os.path.join(DEBUG_FILE_DIR, "lowering_passes_visualization")
+ # path = os.path.join(DEBUG_FILE_DIR, "lowering_passes_visualization")
- ATEN_PRE_LOWERING_PASSES.insert_debug_pass_before(self.cfg.capture_fx_graph_before)
- ATEN_POST_LOWERING_PASSES.insert_debug_pass_before(self.cfg.capture_fx_graph_before)
+ ATEN_PRE_LOWERING_PASSES.insert_debug_pass_before(
+ self.cfg.capture_fx_graph_before
+ )
+ ATEN_POST_LOWERING_PASSES.insert_debug_pass_before(
+ self.cfg.capture_fx_graph_before
+ )
- ATEN_PRE_LOWERING_PASSES.insert_debug_pass_after(self.cfg.capture_fx_graph_after)
- ATEN_POST_LOWERING_PASSES.insert_debug_pass_after(self.cfg.capture_fx_graph_after)
+ ATEN_PRE_LOWERING_PASSES.insert_debug_pass_after(
+ self.cfg.capture_fx_graph_after
+ )
+ ATEN_POST_LOWERING_PASSES.insert_debug_pass_after(
+ self.cfg.capture_fx_graph_after
+ )
self._context_stack = contextlib.ExitStack()
for f in _DEBUG_ENABLED_FUNCS:
f.__kwdefaults__["_debugger_settings"] = self.cfg
[
self._context_stack.enter_context(
mock.patch.object(
c,
- '__init__',
- functools.partialmethod(
- c.__init__,
- _debugger_settings=self.cfg
- )
+ "__init__",
+ functools.partialmethod(c.__init__, _debugger_settings=self.cfg),
)
- ) for c in _DEBUG_ENABLED_CLS
+ )
+ for c in _DEBUG_ENABLED_CLS
]
def __exit__(self, exc_type: Any, exc_value: Any, exc_tb: Any) -> None:
- #dictConfig(self.get_default_config())
+ # dictConfig(self.get_default_config())
torch.ops.tensorrt.set_logging_level(self.rt_level)
ATEN_PRE_LOWERING_PASSES.passes, ATEN_POST_LOWERING_PASSES.passes = (
self.old_pre_passes,
self.old_post_passes,
@@ -80,12 +89,10 @@
for f in _DEBUG_ENABLED_FUNCS:
f.__kwdefaults__["_debugger_settings"] = None
self._context_stack.close()
-
-
# def get_config(self) -> dict[str, Any]:
# config = {
# "version": 1,
# "disable_existing_loggers": False,
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_supports_debugger.py 2025-06-03 04:51:20.323103+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/_supports_debugger.py 2025-06-03 04:51:43.226752+00:00
@@ -4,12 +4,14 @@
_DEBUG_ENABLED_FUNCS = []
_DEBUG_ENABLED_CLS = []
+
def fn_supports_debugger(func):
_DEBUG_ENABLED_FUNCS.append(func)
return func
-def cls_supports_debugger(cls: Type[T]) -> Type[T]:
+
+def cls_supports_debugger(cls: Type[T]) -> Type[T]:
_DEBUG_ENABLED_CLS.append(cls)
return cls
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTBuilderMonitor.py 2025-06-03 04:51:20.324103+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTBuilderMonitor.py 2025-06-03 04:51:43.415699+00:00
@@ -90,10 +90,11 @@
clear_line()
print()
move_cursor_up(len(self._active_phases) + blank_lines)
sys.stdout.flush()
+
# try:
# from rich.progress import BarColumn, Progress, TaskID, TextColumn, TimeElapsedColumn
# class _RichMonitor(trt.IProgressMonitor): # type: ignore
# def __init__(self, engine_name: str = "") -> None:
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py 2025-06-03 04:51:20.324103+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py 2025-06-03 04:51:44.112858+00:00
@@ -69,10 +69,11 @@
input_names: Sequence[str]
output_names: Sequence[str]
weight_name_map: Optional[dict[Any, Any]]
requires_output_allocator: bool
+
@cls_supports_debugger
class TRTInterpreter(torch.fx.Interpreter): # type: ignore[misc]
def __init__(
self,
module: torch.fx.GraphModule,
@@ -207,11 +208,14 @@
algorithm_selector: Optional[trt.IAlgorithmSelector] = None,
tactic_sources: Optional[int] = None,
) -> trt.IBuilderConfig:
builder_config = self.builder.create_builder_config()
- if self._debugger_settings is not None and self._debugger_settings.engine_builder_monitor:
+ if (
+ self._debugger_settings is not None
+ and self._debugger_settings.engine_builder_monitor
+ ):
builder_config.progress_monitor = TRTBulderMonitor()
if self.compilation_settings.workspace_size != 0:
builder_config.set_memory_pool_limit(
trt.MemoryPoolType.WORKSPACE, self.compilation_settings.workspace_size
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/remove_assert_nodes.py 2025-06-03 04:51:20.328103+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/remove_assert_nodes.py 2025-06-03 04:51:44.527302+00:00
@@ -9,16 +9,23 @@
clean_up_graph_after_modifications,
)
logger = logging.getLogger(__name__)
+
@fn_supports_debugger
def remove_assert_nodes(
- gm: torch.fx.GraphModule, settings: CompilationSettings, *, _debugger_settings: Optional[DebuggerConfig]=None
+ gm: torch.fx.GraphModule,
+ settings: CompilationSettings,
+ *,
+ _debugger_settings: Optional[DebuggerConfig] = None,
) -> torch.fx.GraphModule:
"""Remove assert_scalar ops in the graph"""
- if _debugger_settings is not None and _debugger_settings.break_in_remove_assert_nodes:
+ if (
+ _debugger_settings is not None
+ and _debugger_settings.break_in_remove_assert_nodes
+ ):
breakpoint()
count = 0
for node in gm.graph.nodes:
if (
node.target == torch.ops.aten._assert_scalar.default
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR demonstrates how we can use a context manager to automatically patch the compiler to toggle debugging settings scattered through the codebase. It operates through 3 main contact points for developers, a optional kwarg, a settings struct and a decorator to declare a component as part of the debugging system.
This config will then be available to be stored as a attr in the class or to be used as part of the function.
Any specific settings should be added to
DebuggerConfig. This becomes the user facing API for theDebuggerclass. The developer can use this setting to manage debugging behavior.Users will set the config value using a context manager:
Here
engine_builder_monitoris a feature ofTRTInterpreterandbreak_in_remove_assert_nodesis a feature of theremove_assert_nodespass.Both these components register as part of the debugging system. When the context is entered, for functions, the default kwargs are edited to contain the
DebuggerConfig. For classes, the class is mocked with the argument filled in the constructor. On exit, these modifications are cleaned up.In this PR (which needs to be cleaned), the example
debugging_example.pytriggers both the TRTInterpreter setting to turn on the builder monitor and the lowering pass setting. Breakpoints are set to verify behavior. Correct execution means that on the first compilation two breakpoints are hit, one in the lowering pass and one in the TRTBuilderMonitor. Then on the second compilation, no breakpoints are hit and the script completes.Fixes # (issue)
Type of change
Please delete options that are not relevant and/or add your own.
Checklist: