Note
Click here to download the full example code
(beta) Using TORCH_LOGS python API with torch.compile¶
Created On: Jan 24, 2024 | Last Updated: Jan 31, 2024 | Last Verified: Nov 05, 2024
Author: Michael Lazos
import logging
This tutorial introduces the TORCH_LOGS environment variable, as well as the Python API, and
demonstrates how to apply it to observe the phases of torch.compile.
Note
This tutorial requires PyTorch 2.2.0 or later.
Setup¶
In this example, we’ll set up a simple Python function which performs an elementwise
add and observe the compilation process with TORCH_LOGS Python API.
Note
There is also an environment variable TORCH_LOGS, which can be used to
change logging settings at the command line. The equivalent environment
variable setting is shown for each example.
import torch
# exit cleanly if we are on a device that doesn't support torch.compile
if torch.cuda.get_device_capability() < (7, 0):
print("Skipping because torch.compile is not supported on this device.")
else:
@torch.compile()
def fn(x, y):
z = x + y
return z + 2
inputs = (torch.ones(2, 2, device="cuda"), torch.zeros(2, 2, device="cuda"))
# print separator and reset dynamo
# between each example
def separator(name):
print(f"==================={name}=========================")
torch._dynamo.reset()
separator("Dynamo Tracing")
# View dynamo tracing
# TORCH_LOGS="+dynamo"
torch._logging.set_logs(dynamo=logging.DEBUG)
fn(*inputs)
separator("Traced Graph")
# View traced graph
# TORCH_LOGS="graph"
torch._logging.set_logs(graph=True)
fn(*inputs)
separator("Fusion Decisions")
# View fusion decisions
# TORCH_LOGS="fusion"
torch._logging.set_logs(fusion=True)
fn(*inputs)
separator("Output Code")
# View output code generated by inductor
# TORCH_LOGS="output_code"
torch._logging.set_logs(output_code=True)
fn(*inputs)
separator("")
===================Dynamo Tracing=========================
V0418 18:40:20.524000 634 torch/_dynamo/convert_frame.py:1345] skipping: _is_skip_guard_eval_unsafe_stance (reason: in skipfiles, file: /usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py)
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] torchdynamo start compiling fn /var/lib/workspace/recipes_source/torch_logs.py:39, stack (elided 5 frames):
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/bin/sphinx-build", line 8, in <module>
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] sys.exit(main())
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/cmd/build.py", line 288, in main
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] return make_main(argv)
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/cmd/build.py", line 193, in make_main
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] return make_mode.run_make_mode(argv[1:])
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/cmd/make_mode.py", line 160, in run_make_mode
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] return make.run_generic_build(args[0])
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/cmd/make_mode.py", line 148, in run_generic_build
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] return build_main(args + opts)
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/cmd/build.py", line 272, in build_main
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] app = Sphinx(args.sourcedir, args.confdir, args.outputdir,
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/application.py", line 256, in __init__
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] self._init_builder()
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/application.py", line 314, in _init_builder
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] self.events.emit('builder-inited')
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx/events.py", line 94, in emit
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] results.append(listener.handler(self.app, *args))
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_gallery.py", line 491, in generate_gallery_rst
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] ) = generate_dir_rst(
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_rst.py", line 431, in generate_dir_rst
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] intro, title, cost = generate_file_rst(
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_rst.py", line 1027, in generate_file_rst
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] output_blocks, time_elapsed = execute_script(script_blocks,
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_rst.py", line 945, in execute_script
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] output_blocks.append(execute_code_block(
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_rst.py", line 810, in execute_code_block
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] is_last_expr, mem_max = _exec_and_get_memory(
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_rst.py", line 676, in _exec_and_get_memory
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] mem_max, _ = gallery_conf['call_memory'](
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_gallery.py", line 223, in call_memory
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] return 0., func()
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/usr/local/lib/python3.10/dist-packages/sphinx_gallery/gen_rst.py", line 600, in __call__
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] exec(self.code, self.fake_main.__dict__)
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] File "/var/lib/workspace/recipes_source/torch_logs.py", line 59, in <module>
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0] fn(*inputs)
V0418 18:40:20.525000 634 torch/_dynamo/convert_frame.py:930] [0/0]
I0418 18:40:20.527000 634 torch/_dynamo/symbolic_convert.py:2706] [0/0] Step 1: torchdynamo start tracing fn /var/lib/workspace/recipes_source/torch_logs.py:39
I0418 18:40:20.527000 634 torch/fx/experimental/symbolic_shapes.py:3192] [0/0] create_env
V0418 18:40:20.531000 634 torch/_dynamo/symbolic_convert.py:932] [0/0] [__trace_source] TRACE starts_line /var/lib/workspace/recipes_source/torch_logs.py:41 in fn (fn)
V0418 18:40:20.531000 634 torch/_dynamo/symbolic_convert.py:932] [0/0] [__trace_source] z = x + y
V0418 18:40:20.532000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE LOAD_FAST x []
V0418 18:40:20.532000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE LOAD_FAST y [LazyVariableTracker()]
V0418 18:40:20.532000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE BINARY_ADD None [LazyVariableTracker(), LazyVariableTracker()]
V0418 18:40:20.533000 634 torch/_dynamo/variables/builder.py:2853] [0/0] wrap_to_fake L['x'] (2, 2) StatefulSymbolicContext(dynamic_sizes=[<DimDynamic.STATIC: 2>, <DimDynamic.STATIC: 2>], dynamic_strides=[<DimDynamic.INFER_STRIDE: 4>, <DimDynamic.INFER_STRIDE: 4>], constraint_sizes=[None, None], constraint_strides=[None, None], view_base_context=None, tensor_source=LocalSource(local_name='x', is_input=True, is_derefed_cell_contents=False), shape_env_to_source_to_symbol_cache={}) <class 'torch.Tensor'>
V0418 18:40:20.534000 634 torch/_dynamo/output_graph.py:2156] [0/0] create_graph_input L_x_ L['x'] FakeTensor(..., device='cuda:0', size=(2, 2)) at debug_level 0 before=False
V0418 18:40:20.535000 634 torch/_dynamo/variables/builder.py:2853] [0/0] wrap_to_fake L['y'] (2, 2) StatefulSymbolicContext(dynamic_sizes=[<DimDynamic.STATIC: 2>, <DimDynamic.STATIC: 2>], dynamic_strides=[<DimDynamic.INFER_STRIDE: 4>, <DimDynamic.INFER_STRIDE: 4>], constraint_sizes=[None, None], constraint_strides=[None, None], view_base_context=None, tensor_source=LocalSource(local_name='y', is_input=True, is_derefed_cell_contents=False), shape_env_to_source_to_symbol_cache={}) <class 'torch.Tensor'>
V0418 18:40:20.536000 634 torch/_dynamo/output_graph.py:2156] [0/0] create_graph_input L_y_ L['y'] FakeTensor(..., device='cuda:0', size=(2, 2)) at debug_level 0 before=False
V0418 18:40:20.538000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE STORE_FAST z [TensorVariable()]
V0418 18:40:20.538000 634 torch/_dynamo/symbolic_convert.py:932] [0/0] [__trace_source] TRACE starts_line /var/lib/workspace/recipes_source/torch_logs.py:42 in fn (fn)
V0418 18:40:20.538000 634 torch/_dynamo/symbolic_convert.py:932] [0/0] [__trace_source] return z + 2
V0418 18:40:20.538000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE LOAD_FAST z []
V0418 18:40:20.539000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE LOAD_CONST 2 [TensorVariable()]
V0418 18:40:20.539000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE BINARY_ADD None [TensorVariable(), ConstantVariable(int: 2)]
V0418 18:40:20.540000 634 torch/_dynamo/symbolic_convert.py:955] [0/0] [__trace_bytecode] TRACE RETURN_VALUE None [TensorVariable()]
I0418 18:40:20.540000 634 torch/_dynamo/symbolic_convert.py:3028] [0/0] Step 1: torchdynamo done tracing fn (RETURN_VALUE)
V0418 18:40:20.541000 634 torch/_dynamo/symbolic_convert.py:3032] [0/0] RETURN_VALUE triggered compile
V0418 18:40:20.541000 634 torch/_dynamo/output_graph.py:972] [0/0] COMPILING GRAPH due to GraphCompileReason(reason='return_value', user_stack=[<FrameSummary file /var/lib/workspace/recipes_source/torch_logs.py, line 42 in fn>], graph_break=False)
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] TRACED GRAPH
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] ===== __compiled_fn_1 =====
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] /usr/local/lib/python3.10/dist-packages/torch/fx/_lazy_graph_module.py class GraphModule(torch.nn.Module):
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] def forward(self, L_x_: "f32[2, 2][2, 1]cuda:0", L_y_: "f32[2, 2][2, 1]cuda:0"):
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] l_x_ = L_x_
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] l_y_ = L_y_
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code]
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] # File: /var/lib/workspace/recipes_source/torch_logs.py:41 in fn, code: z = x + y
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] z: "f32[2, 2][2, 1]cuda:0" = l_x_ + l_y_; l_x_ = l_y_ = None
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code]
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] # File: /var/lib/workspace/recipes_source/torch_logs.py:42 in fn, code: return z + 2
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] add_1: "f32[2, 2][2, 1]cuda:0" = z + 2; z = None
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code] return (add_1,)
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code]
V0418 18:40:20.542000 634 torch/_dynamo/output_graph.py:1353] [0/0] [__graph_code]
I0418 18:40:20.544000 634 torch/_dynamo/output_graph.py:1458] [0/0] Step 2: calling compiler function inductor
I0418 18:40:21.789000 634 torch/fx/experimental/symbolic_shapes.py:4547] [0/0] produce_guards
I0418 18:40:21.793000 634 torch/_dynamo/output_graph.py:1463] [0/0] Step 2: done compiler function inductor
I0418 18:40:21.795000 634 torch/fx/experimental/symbolic_shapes.py:4547] [0/0] produce_guards
V0418 18:40:21.796000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['x'].size()[0] 2 None
V0418 18:40:21.796000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['x'].size()[1] 2 None
V0418 18:40:21.796000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['x'].stride()[0] 2 None
V0418 18:40:21.796000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['x'].stride()[1] 1 None
V0418 18:40:21.797000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['x'].storage_offset() 0 None
V0418 18:40:21.797000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['y'].size()[0] 2 None
V0418 18:40:21.797000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['y'].size()[1] 2 None
V0418 18:40:21.797000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['y'].stride()[0] 2 None
V0418 18:40:21.797000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['y'].stride()[1] 1 None
V0418 18:40:21.798000 634 torch/fx/experimental/symbolic_shapes.py:4755] [0/0] track_symint L['y'].storage_offset() 0 None
V0418 18:40:21.798000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['x'].size()[0] == 2
V0418 18:40:21.798000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['x'].size()[1] == 2
V0418 18:40:21.799000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['x'].stride()[0] == 2
V0418 18:40:21.799000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['x'].stride()[1] == 1
V0418 18:40:21.799000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['x'].storage_offset() == 0
V0418 18:40:21.799000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['y'].size()[0] == 2
V0418 18:40:21.800000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['y'].size()[1] == 2
V0418 18:40:21.800000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['y'].stride()[0] == 2
V0418 18:40:21.800000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['y'].stride()[1] == 1
V0418 18:40:21.800000 634 torch/fx/experimental/symbolic_shapes.py:4958] [0/0] Skipping guard L['y'].storage_offset() == 0
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2364] [0/0] [__guards] GUARDS:
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards]
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] TREE_GUARD_MANAGER:
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] +- RootGuardManager
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | +- DEFAULT_DEVICE: utils_device.CURRENT_DEVICE == None # _dynamo/output_graph.py:493 in init_ambient_guards
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | +- GLOBAL_STATE: ___check_global_state()
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | +- TORCH_FUNCTION_MODE_STACK: ___check_torch_function_mode_stack()
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | +- GuardManager: source=L['x'], accessed_by=DictGetItemGuardAccessor('x')
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | | +- TENSOR_MATCH: check_tensor(L['x'], Tensor, DispatchKeySet(CUDA, BackendSelect, ADInplaceOrView, AutogradCUDA), torch.float32, device=0, requires_grad=False, size=[2, 2], stride=[2, 1]) # z = x + y # ar/lib/workspace/recipes_source/torch_logs.py:41 in fn
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | | +- NO_HASATTR: hasattr(L['x'], '_dynamo_dynamic_indices') == False # z = x + y # ar/lib/workspace/recipes_source/torch_logs.py:41 in fn
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | | +- NO_TENSOR_ALIASING: check_no_aliasing(L['x'], L['y'])
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | +- GuardManager: source=L['y'], accessed_by=DictGetItemGuardAccessor('y')
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | | +- TENSOR_MATCH: check_tensor(L['y'], Tensor, DispatchKeySet(CUDA, BackendSelect, ADInplaceOrView, AutogradCUDA), torch.float32, device=0, requires_grad=False, size=[2, 2], stride=[2, 1]) # z = x + y # ar/lib/workspace/recipes_source/torch_logs.py:41 in fn
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | | +- NO_HASATTR: hasattr(L['y'], '_dynamo_dynamic_indices') == False # z = x + y # ar/lib/workspace/recipes_source/torch_logs.py:41 in fn
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards] | | +- NO_TENSOR_ALIASING
V0418 18:40:21.801000 634 torch/_dynamo/guards.py:2321] [0/0] [__guards]
V0418 18:40:22.802000 634 torch/_dynamo/guards.py:2346] [0/0] [__guards] Guard eval latency = 0.90 us
I0418 18:40:22.803000 634 torch/_dynamo/pgo.py:636] [0/0] put_code_state: no cache key, skipping
V0418 18:40:22.807000 634 torch/_dynamo/convert_frame.py:1345] skipping: _fn (reason: in skipfiles, file: /usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py)
V0418 18:40:22.808000 634 torch/_dynamo/convert_frame.py:1345] skipping: _callback_from_stance (reason: in skipfiles, file: /usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py)
V0418 18:40:22.808000 634 torch/_dynamo/convert_frame.py:1345] skipping: _maybe_set_eval_frame (reason: in skipfiles, file: /usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py)
V0418 18:40:22.808000 634 torch/_dynamo/convert_frame.py:1345] skipping: justknobs_check (reason: in skipfiles, file: /usr/local/lib/python3.10/dist-packages/torch/_utils_internal.py)
===================Traced Graph=========================
I0418 18:40:22.809000 634 torch/_dynamo/__init__.py:99] torch._dynamo.reset
I0418 18:40:22.809000 634 torch/_dynamo/__init__.py:132] torch._dynamo.reset_code_caches
===================Fusion Decisions=========================
===================Output Code=========================
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1116] [0/0] [__output_code] Output code written to: /tmp/torchinductor_ci-user/kq/ckqum5d54uw5vrn4ovqu3ozxvleh2tky6bhsgwcpfvkh2e2266p2.py
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] Output code:
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # AOT ID: ['1_inference']
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from ctypes import c_void_p, c_long, c_int
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import torch
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import math
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import random
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import os
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import tempfile
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from math import inf, nan
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.hooks import run_intermediate_hooks
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.utils import maybe_profile
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.codegen.memory_planning import _align as align
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch import device, empty_strided
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.async_compile import AsyncCompile
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.select_algorithm import extern_kernels
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.codegen.multi_kernel import MultiKernelCall
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import triton
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import triton.language as tl
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.runtime.triton_heuristics import (
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] grid,
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] split_scan_grid,
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] grid_combo_kernels,
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] start_graph,
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] end_graph,
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] cooperative_reduction_grid,
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] )
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._C import _cuda_getCurrentRawStream as get_raw_stream
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._C import _cuda_getCurrentRawStream as get_raw_stream
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] aten = torch.ops.aten
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] inductor_ops = torch.ops.inductor
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] _quantized = torch.ops._quantized
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] assert_size_stride = torch._C._dynamo.guards.assert_size_stride
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] alloc_from_pool = torch.ops.inductor._alloc_from_pool
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] async_compile = AsyncCompile()
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] empty_strided_p2p = torch._C._distributed_c10d._SymmetricMemory.empty_strided_p2p
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # kernel path: /tmp/torchinductor_ci-user/ea/ceamkuk7na23fw4f7lhqhrqai3i7ypqjiqxrfg3gsladwvs7sqfl.py
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # Topologically Sorted Source Nodes: [z, add_1], Original ATen: [aten.add]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # Source node to ATen node mapping:
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # add_1 => add_1
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # z => add
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # Graph fragment:
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # %add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arg0_1, %arg1_1), kwargs = {})
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # %add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%add, 2), kwargs = {})
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] triton_poi_fused_add_0 = async_compile.triton('triton_poi_fused_add_0', '''
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import triton
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] import triton.language as tl
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from triton.compiler.compiler import AttrsDescriptor
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.runtime import triton_helpers, triton_heuristics
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, DeviceProperties
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] triton_helpers.set_driver_to_gpu()
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] @triton_heuristics.pointwise(
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] size_hints={'x': 4},
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] filename=__file__,
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] triton_meta={'signature': {'in_ptr0': '*fp32', 'in_ptr1': '*fp32', 'out_ptr0': '*fp32', 'xnumel': 'i32'}, 'device': DeviceProperties(type='cuda', index=0, multi_processor_count=80, cc=86, major=8, regs_per_multiprocessor=65536, max_threads_per_multi_processor=1536, warp_size=32), 'constants': {}, 'configs': [AttrsDescriptor.from_dict({'arg_properties': {'tt.divisibility': (0, 1, 2), 'tt.equal_to': ()}, 'cls': 'AttrsDescriptor'})]},
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_add_0', 'mutated_arg_names': [], 'optimize_mem': True, 'no_x_dim': False, 'num_load': 2, 'num_reduction': 0, 'backend_hash': '0E91B58DAB54C915AAF8467E3EDB6871F6D05685FF049BBEEDA70C789216121A', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False},
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] min_elem_per_thread=0
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] )
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] @triton.jit
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] def triton_poi_fused_add_0(in_ptr0, in_ptr1, out_ptr0, xnumel, XBLOCK : tl.constexpr):
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] xnumel = 4
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] xoffset = tl.program_id(0) * XBLOCK
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] xindex = xoffset + tl.arange(0, XBLOCK)[:]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] xmask = xindex < xnumel
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] x0 = xindex
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] tmp0 = tl.load(in_ptr0 + (x0), xmask)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] tmp1 = tl.load(in_ptr1 + (x0), xmask)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] tmp2 = tmp0 + tmp1
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] tmp3 = 2.0
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] tmp4 = tmp2 + tmp3
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] tl.store(out_ptr0 + (x0), tmp4, xmask)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] ''', device_str='cuda')
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] async_compile.wait(globals())
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] del async_compile
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] def call(args):
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] arg0_1, arg1_1 = args
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] args.clear()
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] assert_size_stride(arg0_1, (2, 2), (2, 1))
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] assert_size_stride(arg1_1, (2, 2), (2, 1))
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] with torch.cuda._DeviceGuard(0):
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] torch.cuda.set_device(0)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] buf0 = empty_strided_cuda((2, 2), (2, 1), torch.float32)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] # Topologically Sorted Source Nodes: [z, add_1], Original ATen: [aten.add]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] stream0 = get_raw_stream(0)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] triton_poi_fused_add_0.run(arg0_1, arg1_1, buf0, 4, grid=grid(4), stream=stream0)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] del arg0_1
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] del arg1_1
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] return (buf0, )
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] def benchmark_compiled_module(times=10, repeat=10):
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._dynamo.testing import rand_strided
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.utils import print_performance
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] arg0_1 = rand_strided((2, 2), (2, 1), device='cuda:0', dtype=torch.float32)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] arg1_1 = rand_strided((2, 2), (2, 1), device='cuda:0', dtype=torch.float32)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] fn = lambda: call([arg0_1, arg1_1])
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] return print_performance(fn, times=times, repeat=repeat)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] if __name__ == "__main__":
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] from torch._inductor.wrapper_benchmark import compiled_module_main
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code] compiled_module_main('None', benchmark_compiled_module)
V0418 18:40:23.163000 634 torch/_inductor/codecache.py:1117] [0/0] [__output_code]
============================================
Conclusion¶
In this tutorial we introduced the TORCH_LOGS environment variable and python API by experimenting with a small number of the available logging options. To view descriptions of all available options, run any python script which imports torch and set TORCH_LOGS to “help”.
Alternatively, you can view the torch._logging documentation to see descriptions of all available logging options.
For more information on torch.compile, see the torch.compile tutorial.
Total running time of the script: ( 0 minutes 3.037 seconds)
