IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[MO][PT FE] POC Integrate PT FE into MO convert_model (#15197)

Browse Source 
* [WIP]: poc: integrate pt fe to mo.convert_model

* temporaly resolve names using signature inspection

* fix logic

* move to moc frontend

* add rank and data type setting

* add input signature

* apply comments

* move after convert processing

* support work without example_input, provide parameters description

* ignore extensions on fe side

* upd tests

* more tests

* add notes about extensions support and removing onnx path

* remove debug prints

* refactoring

* update BOM
This commit is contained in:
Ekaterina Aidova
2023-02-15 11:19:15 +04:00
committed by GitHub
parent 36cb32a8f5
commit 38943434b6
8 changed files with 491 additions and 260 deletions
Download Patch File Download Diff File Expand all files Collapse all files

171
tests/layer_tests/mo_python_api_tests/test_mo_convert_pytorch.py
View File

@@ -8,7 +8,7 @@ import numpy as np
import openvino.runtime as ov
import pytest
import torch
from openvino.runtime import PartialShape, Dimension, Model
from openvino.runtime import PartialShape, Dimension, Model, Type
from common.mo_convert_test_class import CommonMOConvertTest
@@ -84,7 +84,7 @@ def create_pytorch_nn_module_case1(tmp_dir):
sample_input = sample_input1, sample_input2
return pt_model, ref_model, {'input_shape': [PartialShape([-1, 3, -1, -1]), PartialShape([-1, 3, -1, -1])],
'example_input': sample_input}
'example_input': sample_input, "use_legacy_frontend": True}
def create_pytorch_nn_module_case2(tmp_dir):
@@ -96,7 +96,7 @@ def create_pytorch_nn_module_case2(tmp_dir):
sample_input = sample_input1, sample_input2
return pt_model, ref_model, {'input_shape': ["[?,3,?,?]", PartialShape([-1, 3, -1, -1])],
'example_input': sample_input, 'onnx_opset_version': 11}
'example_input': sample_input, 'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_case3(tmp_dir):
@@ -107,7 +107,7 @@ def create_pytorch_nn_module_case3(tmp_dir):
sample_input2 = torch.zeros(1, 3, 10, 10)
sample_input = tuple([sample_input1, sample_input2])
return pt_model, ref_model, {'input_shape': "[?,3,?,?],[?,3,?,?]", 'example_input': sample_input}
return pt_model, ref_model, {'input_shape': "[?,3,?,?],[?,3,?,?]", 'example_input': sample_input, "use_legacy_frontend": True}
def create_pytorch_nn_module_case4(tmp_dir):
@@ -117,7 +117,7 @@ def create_pytorch_nn_module_case4(tmp_dir):
ref_model = make_ref_pt_model_one_input([1, 3, 10, 10])
return pt_model, ref_model, {'example_input': sample_input}
return pt_model, ref_model, {'example_input': sample_input, "use_legacy_frontend": True}
def create_pytorch_nn_module_case5(tmp_dir):
@@ -127,7 +127,7 @@ def create_pytorch_nn_module_case5(tmp_dir):
sample_input = torch.zeros(3, 3, 10, 10)
return pt_model, ref_model, {'example_input': sample_input,
'input_shape': inp_shape}
'input_shape': inp_shape, "use_legacy_frontend": True}
def create_pytorch_nn_module_case6(tmp_dir):
@@ -135,14 +135,14 @@ def create_pytorch_nn_module_case6(tmp_dir):
shape = PartialShape([1, 3, Dimension(2, -1), Dimension(-1, 10)])
ref_model = make_ref_pt_model_one_input(shape)
return pt_model, ref_model, {'input_shape': shape}
return pt_model, ref_model, {'input_shape': shape, "use_legacy_frontend": True}
def create_pytorch_nn_module_torch_size(tmp_dir):
pt_model = make_pt_model_one_input()
ref_model = make_ref_pt_model_one_input([1, 3, 2, 10])
return pt_model, ref_model, {'input_shape': torch.Size([1, 3, 2, 10])}
return pt_model, ref_model, {'input_shape': torch.Size([1, 3, 2, 10]), "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_input_int32(tmp_dir):
@@ -154,7 +154,7 @@ def create_pytorch_nn_module_sample_input_int32(tmp_dir):
ref_model = make_ref_pt_model_one_input(shape, dtype=numpy.int32)
return pt_model, ref_model, {'example_input': sample_input,
'input_shape': shape}
'input_shape': shape, "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_input_int32_two_inputs(tmp_dir):
@@ -167,7 +167,7 @@ def create_pytorch_nn_module_sample_input_int32_two_inputs(tmp_dir):
ref_model = make_ref_pt_model_two_inputs([PartialShape([-1, 3, -1, -1]), inp_shapes[1]], dtype=np.int32)
return pt_model, ref_model, {'input_shape': inp_shapes,
'example_input': sample_input, 'onnx_opset_version': 11}
'example_input': sample_input, 'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_compare_convert_paths_case1(tmp_dir):
@@ -179,7 +179,7 @@ def create_pytorch_nn_module_compare_convert_paths_case1(tmp_dir):
torch.onnx.export(pt_model, sample_input, onnx_model_path, opset_version=16)
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': sample_input, 'onnx_opset_version': 16}
return pt_model, ref_model, {'example_input': sample_input, 'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_compare_convert_paths_case2(tmp_dir):
@@ -193,7 +193,9 @@ def create_pytorch_nn_module_compare_convert_paths_case2(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': sample_input,
'input_shape': [1, 3, 10, 10],
'onnx_opset_version': 16}
'onnx_opset_version': 16,
"use_legacy_frontend": True
}
def create_pytorch_nn_module_compare_convert_paths_case3(tmp_dir):
@@ -206,7 +208,7 @@ def create_pytorch_nn_module_compare_convert_paths_case3(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'input_shape': [1, 3, 10, 10],
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_compare_convert_paths_case4(tmp_dir):
@@ -222,7 +224,7 @@ def create_pytorch_nn_module_compare_convert_paths_case4(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': sample_input, 'onnx_opset_version': 16}
return pt_model, ref_model, {'example_input': sample_input, 'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_compare_convert_paths_case5(tmp_dir):
@@ -240,7 +242,7 @@ def create_pytorch_nn_module_compare_convert_paths_case5(tmp_dir):
return pt_model, ref_model, {'example_input': sample_input,
'input_shape': [torch.Size([1, 3, 10, 10]), PartialShape([1, 3, 10, 10])],
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_compare_convert_paths_case6(tmp_dir):
@@ -257,7 +259,7 @@ def create_pytorch_nn_module_compare_convert_paths_case6(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'input_shape': [torch.Size([1, 3, 10, 10]), torch.Size([1, 3, 10, 10])],
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_jit_script_module(tmp_dir):
@@ -267,7 +269,7 @@ def create_pytorch_jit_script_module(tmp_dir):
scripted_model = torch.jit.script(net)
model_ref = make_ref_pt_model_two_inputs([1, 3, 5, 5])
return scripted_model, model_ref, {'input_shape': [PartialShape([1, 3, 5, 5]), PartialShape([1, 3, 5, 5])]}
return scripted_model, model_ref, {'input_shape': [PartialShape([1, 3, 5, 5]), PartialShape([1, 3, 5, 5])], "use_legacy_frontend": True}
def create_pytorch_jit_script_function(tmp_dir):
@@ -279,7 +281,7 @@ def create_pytorch_jit_script_function(tmp_dir):
inp_shape = PartialShape([Dimension(1, -1), Dimension(-1, 5), 10])
ref_model = make_ref_pt_model_two_inputs(inp_shape)
return scripted_fn, ref_model, {'input_shape': [inp_shape, inp_shape]}
return scripted_fn, ref_model, {'input_shape': [inp_shape, inp_shape], "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_input_numpy(tmp_dir):
@@ -293,7 +295,7 @@ def create_pytorch_nn_module_sample_input_numpy(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': example_inputs,
'input_shape': [1, 3, 10, 10],
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_input_dict(tmp_dir):
@@ -306,7 +308,7 @@ def create_pytorch_nn_module_sample_input_dict(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': example_inputs,
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_input_dict_two_inputs(tmp_dir):
@@ -321,7 +323,7 @@ def create_pytorch_nn_module_sample_input_dict_two_inputs(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': example_inputs,
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_list_of_tensors(tmp_dir):
@@ -335,7 +337,7 @@ def create_pytorch_nn_module_sample_list_of_tensors(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': example_inputs,
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_input_ov_host_tensor(tmp_dir):
@@ -350,7 +352,7 @@ def create_pytorch_nn_module_sample_input_ov_host_tensor(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': sample_input,
'input_shape': [1, 3, 10, 10],
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_sample_input_ov_host_tensor_two_inputs(tmp_dir):
@@ -370,7 +372,7 @@ def create_pytorch_nn_module_sample_input_ov_host_tensor_two_inputs(tmp_dir):
ref_model = convert_model(onnx_model_path)
return pt_model, ref_model, {'example_input': sample_input,
'onnx_opset_version': 16}
'onnx_opset_version': 16, "use_legacy_frontend": True}
def create_pytorch_nn_module_layout_list(tmp_dir):
@@ -384,7 +386,7 @@ def create_pytorch_nn_module_layout_list(tmp_dir):
ref_model.inputs[1].node.layout = Layout('nhwc')
return pt_model, ref_model, {'input_shape': [shape, shape], 'layout': ['nchw', Layout('nhwc')],
'onnx_opset_version': 11}
'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_layout_list_case2(tmp_dir):
@@ -398,7 +400,7 @@ def create_pytorch_nn_module_layout_list_case2(tmp_dir):
ref_model.inputs[1].node.layout = Layout('nhwc')
return pt_model, ref_model, {'input_shape': [shape, shape], 'layout': ('nchw', Layout('nhwc')),
'onnx_opset_version': 11}
'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_mean_list(tmp_dir):
@@ -420,7 +422,7 @@ def create_pytorch_nn_module_mean_list(tmp_dir):
ref_model = Model([sigm], parameter_list, "test")
return pt_model, ref_model, {'input_shape': [shape, shape], 'mean_values': [[0, 0, 0], [0, 0, 0]],
'onnx_opset_version': 11}
'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_scale_list(tmp_dir):
@@ -442,14 +444,14 @@ def create_pytorch_nn_module_scale_list(tmp_dir):
ref_model = Model([sigm], parameter_list, "test")
return pt_model, ref_model, {'input_shape': [shape, shape], 'scale_values': [[1, 1, 1], [1, 1, 1]],
'onnx_opset_version': 11}
'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_shapes_list_static(tmp_dir):
pt_model = make_pt_model_two_inputs()
ref_model = make_ref_pt_model_two_inputs([1, 3, 20, 20])
return pt_model, ref_model, {'input_shape': [[1, 3, 20, 20], [1, 3, 20, 20]], 'onnx_opset_version': 11}
return pt_model, ref_model, {'input_shape': [[1, 3, 20, 20], [1, 3, 20, 20]], 'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_shapes_list_dynamic(tmp_dir):
@@ -464,21 +466,118 @@ def create_pytorch_nn_module_shapes_list_dynamic(tmp_dir):
parameter_list = [param1, param2]
ref_model = Model([sigm], parameter_list, "test")
return pt_model, ref_model, {'input_shape': inp_shapes, 'onnx_opset_version': 11}
return pt_model, ref_model, {'input_shape': inp_shapes, 'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_shapes_list_dynamic_single_input(tmp_dir):
pt_model = make_pt_model_one_input()
inp_shapes = [[Dimension(-1), 3, 20, Dimension(20, -1)]]
ref_model = make_ref_pt_model_one_input(inp_shapes[0])
return pt_model, ref_model, {'input_shape': inp_shapes, 'onnx_opset_version': 11}
return pt_model, ref_model, {'input_shape': inp_shapes, 'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_shapes_list_static_single_input(tmp_dir):
pt_model = make_pt_model_one_input()
inp_shapes = [[1, 3, 20, 20]]
ref_model = make_ref_pt_model_one_input(inp_shapes[0])
return pt_model, ref_model, {'input_shape': inp_shapes, 'onnx_opset_version': 11}
return pt_model, ref_model, {'input_shape': inp_shapes, 'onnx_opset_version': 11, "use_legacy_frontend": True}
def create_pytorch_nn_module_convert_pytorch_frontend1(tmp_dir):
pt_model = make_pt_model_one_input()
shape = [-1, -1, -1, -1]
shape = PartialShape(shape)
param = ov.opset10.parameter(shape)
relu = ov.opset10.relu(param)
sigm = ov.opset10.sigmoid(relu)
parameter_list = [param]
ref_model = Model([sigm], parameter_list, "test")
return pt_model, ref_model, {"example_input": torch.zeros((1, 3, 10, 10))}
def create_pytorch_nn_module_convert_pytorch_frontend2(tmp_dir):
pt_model = make_pt_model_one_input()
shape = [-1, -1, -1, -1]
shape = PartialShape(shape)
param = ov.opset10.parameter(shape)
relu = ov.opset10.relu(param)
sigm = ov.opset10.sigmoid(relu)
parameter_list = [param]
ref_model = Model([sigm], parameter_list, "test")
ref_model.input(0).get_node().set_element_type(Type.i32)
ref_model.validate_nodes_and_infer_types()
return pt_model, ref_model, {"example_input": torch.zeros((1, 3, 10, 10), dtype=torch.int32)}
def create_pytorch_nn_module_convert_pytorch_frontend3(tmp_dir):
pt_model = make_pt_model_two_inputs()
shape = [-1, -1, -1, -1]
shape = PartialShape(shape)
param1 = ov.opset10.parameter(shape)
param2 = ov.opset10.parameter(shape)
param2_convert = ov.opset10.convert_like(param2, param1)
add = ov.opset10.add(param1, param2_convert)
relu = ov.opset10.relu(add)
sigm = ov.opset10.sigmoid(relu)
parameter_list = [param1, param2]
ref_model = Model([sigm], parameter_list, "test")
return pt_model, ref_model, {"example_input": [torch.zeros((1, 3, 10, 10)), torch.ones((1, 3, 10, 10))]}
def create_pytorch_nn_module_convert_pytorch_frontend4(tmp_dir):
pt_model = make_pt_model_two_inputs()
shape = [-1, -1, -1, -1]
shape = PartialShape(shape)
param1 = ov.opset10.parameter(shape)
param2 = ov.opset10.parameter(shape)
param2_convert = ov.opset10.convert_like(param2, param1)
add = ov.opset10.add(param1, param2_convert)
relu = ov.opset10.relu(add)
sigm = ov.opset10.sigmoid(relu)
parameter_list = [param1, param2]
ref_model = Model([sigm], parameter_list, "test")
return pt_model, ref_model, {"example_input": {"x": torch.zeros((1, 3, 10, 10)), "y": torch.ones((1, 3, 10, 10))}}
def create_pytorch_jit_script_module_convert_pytorch_frontend(tmp_dir):
import torch
net = make_pt_model_two_inputs()
scripted_model = torch.jit.script(net)
shape = [-1, -1, -1, -1]
shape = PartialShape(shape)
param1 = ov.opset10.parameter(shape)
param2 = ov.opset10.parameter(shape)
param2_convert = ov.opset10.convert_like(param2, param1)
add = ov.opset10.add(param1, param2_convert)
relu = ov.opset10.relu(add)
sigm = ov.opset10.sigmoid(relu)
parameter_list = [param1, param2]
ref_model = Model([sigm], parameter_list, "test")
return scripted_model, ref_model, {"example_input": {"x": torch.zeros((1, 3, 10, 10)), "y": torch.ones((1, 3, 10, 10))}}
def create_pytorch_jit_trace_module_convert_pytorch_frontend(tmp_dir):
import torch
net = make_pt_model_two_inputs()
example_input = [torch.zeros((1, 3, 10, 10)), torch.ones((1, 3, 10, 10))]
scripted_model = torch.jit.trace(net, example_input)
shape = [-1, -1, -1, -1]
shape = PartialShape(shape)
param1 = ov.opset10.parameter(shape)
param2 = ov.opset10.parameter(shape)
param2_convert = ov.opset10.convert_like(param2, param1)
add = ov.opset10.add(param1, param2_convert)
relu = ov.opset10.relu(add)
sigm = ov.opset10.sigmoid(relu)
parameter_list = [param1, param2]
ref_model = Model([sigm], parameter_list, "test")
return scripted_model, ref_model, {"example_input": example_input}
class TestMoConvertPyTorch(CommonMOConvertTest):
@@ -512,7 +611,13 @@ class TestMoConvertPyTorch(CommonMOConvertTest):
create_pytorch_nn_module_shapes_list_static,
create_pytorch_nn_module_shapes_list_dynamic,
create_pytorch_nn_module_shapes_list_dynamic_single_input,
create_pytorch_nn_module_shapes_list_static_single_input
create_pytorch_nn_module_shapes_list_static_single_input,
create_pytorch_nn_module_convert_pytorch_frontend1,
create_pytorch_nn_module_convert_pytorch_frontend2,
create_pytorch_nn_module_convert_pytorch_frontend3,
create_pytorch_nn_module_convert_pytorch_frontend4,
create_pytorch_jit_script_module_convert_pytorch_frontend,
create_pytorch_jit_trace_module_convert_pytorch_frontend
]
@pytest.mark.parametrize("create_model", test_data)