Skip to content

【Fix PIR Unittest】fix some public API and inference UT in pir #66481

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jul 26, 2024
Merged

【Fix PIR Unittest】fix some public API and inference UT in pir #66481

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
50df4af
【Fix PIR Unittest】fix some public API and inference UT in pir
lizhenyun01 Jul 24, 2024
7fcaf10
fix test_mkldnn_conv_gelu_fuse_pass_pir time_out
lizhenyun01 Jul 25, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
54 changes: 54 additions & 0 deletions test/ir/inference/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -307,3 +307,57 @@ if(WITH_GPU AND TENSORRT_FOUND)
endif()
endif()
endif()

if(WITH_GPU AND TENSORRT_FOUND)
if(WITH_ONEDNN)
py_test_modules(test_matmul_scale_fuse_pass_pir MODULES
test_matmul_scale_fuse_pass ENVS FLAGS_enable_pir_api=1)
if(WIN32)
set_tests_properties(test_matmul_scale_fuse_pass_pir PROPERTIES TIMEOUT
300)
else()
py_test_modules(
test_matmul_v2_scale_fuse_pass_pir MODULES
test_matmul_v2_scale_fuse_pass ENVS FLAGS_enable_pir_api=1)
set_tests_properties(test_matmul_scale_fuse_pass_pir PROPERTIES TIMEOUT
60)
set_tests_properties(test_matmul_v2_scale_fuse_pass_pir PROPERTIES TIMEOUT
120)
endif()
if(WIN32 AND WIN_UNITTEST_LEVEL LESS 2)
message(STATUS "Skip tests unrelated to CUDA/TRT")
else()
set(PIR_COVERAGE_MKLDNN_TESTS
test_mkldnn_conv_affine_channel_fuse_pass
test_mkldnn_conv_gelu_fuse_pass
test_mkldnn_conv_hard_sigmoid_fuse_pass
test_mkldnn_conv_hard_swish_fuse_pass
test_mkldnn_conv_mish_fuse_pass
test_mkldnn_conv_transpose_bias_fuse_pass
test_mkldnn_conv3d_op
test_mkldnn_depthwise_conv_pass)
foreach(target ${PIR_COVERAGE_MKLDNN_TESTS})
py_test_modules(${target}_pir MODULES ${target} ENVS
FLAGS_enable_pir_api=1)
set_tests_properties(${target} PROPERTIES LABELS "RUN_TYPE=INFER")
message(STATUS "PIR Copied Test: ${target}_pir in inference test")
endforeach()

set_tests_properties(test_mkldnn_conv_affine_channel_fuse_pass_pir
PROPERTIES TIMEOUT 120)
set_tests_properties(test_mkldnn_conv_gelu_fuse_pass_pir
PROPERTIES TIMEOUT 300)
set_tests_properties(test_mkldnn_conv_hard_sigmoid_fuse_pass_pir
PROPERTIES TIMEOUT 300)
set_tests_properties(test_mkldnn_conv_hard_swish_fuse_pass_pir
PROPERTIES TIMEOUT 300)
set_tests_properties(test_mkldnn_conv_mish_fuse_pass_pir
PROPERTIES TIMEOUT 300)
set_tests_properties(test_mkldnn_conv_transpose_bias_fuse_pass_pir
PROPERTIES TIMEOUT 100)
set_tests_properties(test_mkldnn_conv3d_op_pir PROPERTIES TIMEOUT 300)
set_tests_properties(test_mkldnn_depthwise_conv_pass_pir
PROPERTIES TIMEOUT 120)
endif()
endif()
endif()
25 changes: 13 additions & 12 deletions test/ir/inference/auto_scan_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,18 +132,19 @@ def run_test_config(
"""
Test a single case.
"""
pred_config.set_model_buffer(model, len(model), params, len(params))
predictor = paddle_infer.create_predictor(pred_config)
self.available_passes_in_framework = (
self.available_passes_in_framework
| set(pred_config.pass_builder().all_passes())
)
for name, _ in prog_config.inputs.items():
input_tensor = predictor.get_input_handle(name)
input_tensor.copy_from_cpu(feed_data[name]["data"])
if feed_data[name]["lod"] is not None:
input_tensor.set_lod(feed_data[name]["lod"])
predictor.run()
with paddle.pir_utils.OldIrGuard():
Copy link
Contributor

@YuanRisheng YuanRisheng Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

同下

pred_config.set_model_buffer(model, len(model), params, len(params))
predictor = paddle_infer.create_predictor(pred_config)
self.available_passes_in_framework = (
self.available_passes_in_framework
| set(pred_config.pass_builder().all_passes())
)
for name, _ in prog_config.inputs.items():
input_tensor = predictor.get_input_handle(name)
input_tensor.copy_from_cpu(feed_data[name]["data"])
if feed_data[name]["lod"] is not None:
input_tensor.set_lod(feed_data[name]["lod"])
predictor.run()
result = {}
for out_name, o_name in zip(
prog_config.outputs, predictor.get_output_names()
Expand Down
252 changes: 133 additions & 119 deletions test/ir/inference/program_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -350,133 +350,147 @@ def create_fake_model(program_config):
program_config = copy.deepcopy(program_config)
program_config._cast()
paddle.enable_static()
main_program_desc = core.ProgramDesc()
util_program = base.Program()
main_block_desc = main_program_desc.block(0)

var_desc = main_block_desc.var(b"feed")
var_desc.set_type(core.VarDesc.VarType.FEED_MINIBATCH)
var_desc.set_persistable(True)

index = 0
for name, tensor_config in program_config.inputs.items():
var_desc = main_block_desc.var(name.encode())
var_desc.set_type(core.VarDesc.VarType.LOD_TENSOR)
var_desc.set_dtype(convert_np_dtype_to_proto_type(tensor_config.dtype))
var_desc.set_shape(tensor_config.shape)
var_desc.set_need_check_feed(True)
if tensor_config.lod is not None:
var_desc.set_lod_level(len(tensor_config.lod))
op_desc = main_block_desc._prepend_op()
op_desc.set_type("feed")
op_desc.set_input('X', ["feed"])
op_desc.set_output('Out', [name])
op_desc._set_attr("col", index)
index = index + 1

save_var_map = {}
for name, tensor_config in program_config.weights.items():
var_desc = main_block_desc.var(name.encode())
var_desc.set_type(core.VarDesc.VarType.LOD_TENSOR)
var_desc.set_dtype(convert_np_dtype_to_proto_type(tensor_config.dtype))
var_desc.set_shape(tensor_config.shape)
with paddle.pir_utils.OldIrGuard():
Copy link
Contributor

@YuanRisheng YuanRisheng Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

我们需要修复PIR模式下的问题单测，但使用OldIrGuard的方式相当于切换回了旧IR运行，自然不会报错，未来OldIr会下线，那这个时候这个单测该怎么处理呢？所以我觉得需要搞清楚问题单测是否真正需要在PIR模式下运行，如果测试内容不需要的话，可以将单测迁移至deprecated目录下后续废弃掉

Copy link
Contributor Author

@lizhenyun01 lizhenyun01 Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

目前讨论的方式是对于未来新加入的pir API及pass建议使用新的inference 单测API和开发方式，目前针对于推全中用于旧IR的pass和API暂时以OldIrGuard的方式支持，后续可以考虑再迁移到新的API；至于是否需要废弃可能需要进一步进行讨论分类

Copy link
Contributor

@vivienfanghuagood vivienfanghuagood Jul 26, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

我们需要修复PIR模式下的问题单测，但使用OldIrGuard的方式相当于切换回了旧IR运行，自然不会报错，未来OldIr会下线，那这个时候这个单测该怎么处理呢？所以我觉得需要搞清楚问题单测是否真正需要在PIR模式下运行，如果测试内容不需要的话，可以将单测迁移至deprecated目录下后续废弃掉

这些单测目前不需要在PIR下运行，但仍暂时保留这些单测

main_program_desc = core.ProgramDesc()
# util_program = base.Program()
Copy link
Contributor

@YuanRisheng YuanRisheng Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

无效注释可以删掉

Copy link
Contributor Author

@lizhenyun01 lizhenyun01 Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

done

util_program = paddle.static.Program()
main_block_desc = main_program_desc.block(0)

var_desc = main_block_desc.var(b"feed")
var_desc.set_type(core.VarDesc.VarType.FEED_MINIBATCH)
var_desc.set_persistable(True)

save_var_map[name] = util_program.global_block().create_parameter(
dtype=tensor_config.dtype,
shape=tensor_config.shape,
type=core.VarDesc.VarType.LOD_TENSOR,
name=name,
initializer=paddle.nn.initializer.Assign(tensor_config.data),
)
in_vars = []
for name in sorted(save_var_map.keys()):
in_vars.append(save_var_map[name])

out_var = util_program.global_block().create_var(
type=core.VarDesc.VarType.RAW, name="out_var_0"
)
out_var.desc.set_persistable(True)
util_program.global_block().append_op(
type='save_combine',
inputs={'X': in_vars},
outputs={'Y': out_var},
attrs={'file_path': '', 'save_to_memory': True},
)
for op_config in program_config.ops:
op_desc = main_block_desc.append_op()
op_desc.set_type(op_config.type)
# canonicalize scalar attrs
if OpProtoHolder.instance().has_op_proto(op_config.type):
proto = OpProtoHolder.instance().get_op_proto(op_config.type)
canonicalized_attrs = framework.canonicalize_attrs(
op_config.attrs, proto
index = 0
for name, tensor_config in program_config.inputs.items():
var_desc = main_block_desc.var(name.encode())
var_desc.set_type(core.VarDesc.VarType.LOD_TENSOR)
var_desc.set_dtype(
convert_np_dtype_to_proto_type(tensor_config.dtype)
)
else:
canonicalized_attrs = op_config.attrs

for name, values in op_config.inputs.items():
op_desc.set_input(name, values)
for name, values in canonicalized_attrs.items():
if name == 'sub_block':
sub_block_desc = main_program_desc.append_block(main_block_desc)
values.fill_block_desc(sub_block_desc)
op_desc._set_attr(name, sub_block_desc)
var_desc.set_shape(tensor_config.shape)
var_desc.set_need_check_feed(True)
if tensor_config.lod is not None:
var_desc.set_lod_level(len(tensor_config.lod))
op_desc = main_block_desc._prepend_op()
op_desc.set_type("feed")
op_desc.set_input('X', ["feed"])
op_desc.set_output('Out', [name])
op_desc._set_attr("col", index)
index = index + 1

save_var_map = {}
for name, tensor_config in program_config.weights.items():
var_desc = main_block_desc.var(name.encode())
var_desc.set_type(core.VarDesc.VarType.LOD_TENSOR)
var_desc.set_dtype(
convert_np_dtype_to_proto_type(tensor_config.dtype)
)
var_desc.set_shape(tensor_config.shape)
var_desc.set_persistable(True)

save_var_map[name] = util_program.global_block().create_parameter(
dtype=tensor_config.dtype,
shape=tensor_config.shape,
type=core.VarDesc.VarType.LOD_TENSOR,
name=name,
initializer=paddle.nn.initializer.Assign(tensor_config.data),
)
in_vars = []
for name in sorted(save_var_map.keys()):
in_vars.append(save_var_map[name])

out_var = util_program.global_block().create_var(
type=core.VarDesc.VarType.RAW, name="out_var_0"
)
out_var.desc.set_persistable(True)
util_program.global_block().append_op(
type='save_combine',
inputs={'X': in_vars},
outputs={'Y': out_var},
attrs={'file_path': '', 'save_to_memory': True},
)
for op_config in program_config.ops:
op_desc = main_block_desc.append_op()
op_desc.set_type(op_config.type)
# canonicalize scalar attrs
if OpProtoHolder.instance().has_op_proto(op_config.type):
proto = OpProtoHolder.instance().get_op_proto(op_config.type)
canonicalized_attrs = framework.canonicalize_attrs(
op_config.attrs, proto
)
else:
op_desc._set_attr(name, values)
for name, values in op_config.outputs.items():
op_desc.set_output(name, values)
for v in values:
if main_block_desc.has_var_recursive(v.encode()):
continue
var_desc = main_block_desc.var(v.encode())
var_desc.set_type(core.VarDesc.VarType.LOD_TENSOR)
if (
op_config.outputs_var_type is not None
and v in op_config.outputs_var_type.keys()
):
canonicalized_attrs = op_config.attrs

for name, values in op_config.inputs.items():
op_desc.set_input(name, values)
for name, values in canonicalized_attrs.items():
if name == 'sub_block':
sub_block_desc = main_program_desc.append_block(
main_block_desc
)
values.fill_block_desc(sub_block_desc)
op_desc._set_attr(name, sub_block_desc)
else:
op_desc._set_attr(name, values)
for name, values in op_config.outputs.items():
op_desc.set_output(name, values)
for v in values:
if main_block_desc.has_var_recursive(v.encode()):
continue
var_desc = main_block_desc.var(v.encode())
var_desc.set_type(core.VarDesc.VarType.LOD_TENSOR)
if (
op_config.outputs_var_type[v]
== VarType.LOD_TENSOR_ARRAY
op_config.outputs_var_type is not None
and v in op_config.outputs_var_type.keys()
):
var_desc.set_type(core.VarDesc.VarType.LOD_TENSOR_ARRAY)
elif op_config.outputs_var_type[v] == VarType.STEP_SCOPES:
var_desc.set_type(core.VarDesc.VarType.STEP_SCOPES)
continue
var_desc.set_dtype(convert_np_dtype_to_proto_type(np.float32))
if (
op_config.outputs_dtype is not None
and v in op_config.outputs_dtype.keys()
):
if (
op_config.outputs_var_type[v]
== VarType.LOD_TENSOR_ARRAY
):
var_desc.set_type(
core.VarDesc.VarType.LOD_TENSOR_ARRAY
)
elif (
op_config.outputs_var_type[v] == VarType.STEP_SCOPES
):
var_desc.set_type(core.VarDesc.VarType.STEP_SCOPES)
continue
var_desc.set_dtype(
convert_np_dtype_to_proto_type(
op_config.outputs_dtype[v]
)
convert_np_dtype_to_proto_type(np.float32)
)
if op_config.type not in _OP_WITHOUT_KERNEL_SET:
op_desc.infer_var_type(main_block_desc)
op_desc.infer_shape(main_block_desc)
op_desc.check_attrs()

for index, name in enumerate(program_config.outputs):
var_desc = main_block_desc.var(b"fetch")
var_desc.set_type(core.VarDesc.VarType.FETCH_LIST)
var_desc.set_need_check_feed(True)
op_desc = main_block_desc.append_op()
op_desc.set_type("fetch")
op_desc.set_input('X', [name])
op_desc.set_output('Out', ["fetch"])
op_desc._set_attr("col", index)

model = main_program_desc.serialize_to_string()

util_program._sync_with_cpp()
place = base.CPUPlace()
executor = base.Executor(place)
scope = base.Scope()
with base.scope_guard(scope):
executor.run(util_program)
params = scope.find_var("out_var_0").get_bytes()
if (
op_config.outputs_dtype is not None
and v in op_config.outputs_dtype.keys()
):
var_desc.set_dtype(
convert_np_dtype_to_proto_type(
op_config.outputs_dtype[v]
)
)
if op_config.type not in _OP_WITHOUT_KERNEL_SET:
op_desc.infer_var_type(main_block_desc)
op_desc.infer_shape(main_block_desc)
op_desc.check_attrs()

for index, name in enumerate(program_config.outputs):
var_desc = main_block_desc.var(b"fetch")
var_desc.set_type(core.VarDesc.VarType.FETCH_LIST)
var_desc.set_need_check_feed(True)
op_desc = main_block_desc.append_op()
op_desc.set_type("fetch")
op_desc.set_input('X', [name])
op_desc.set_output('Out', ["fetch"])
op_desc._set_attr("col", index)

model = main_program_desc.serialize_to_string()

util_program._sync_with_cpp()
place = base.CPUPlace()
executor = base.Executor(place)
scope = base.Scope()
with base.scope_guard(scope):
executor.run(util_program)
params = scope.find_var("out_var_0").get_bytes()

return model, params

Expand Down