mindstudio-probe 1.0.1__py3-none-any.whl → 1.0.3__py3-none-any.whl

msprobe/test/pytorch_ut/hook_module/test_api_registry.py

@@ -1,130 +0,0 @@
-import unittest
-from msprobe.pytorch.hook_module.api_registry import ApiRegistry, torch_version_above_2, is_gpu
-
-
-class TestApiRegistry(unittest.TestCase):
-    
-    def test_store_ori_attr(self):
-        class A():
-            a1 = 1
-        class B():
-            a = A()
-            b1 = 1
-            b2 = 2
-        
-        api_list = ["a.a1", "b1", "b2"]
-        expect_output = {"a.a1":1, "b1":1, "b2":2}
-        actual_output = dict()
-        ApiRegistry.store_ori_attr(B, api_list, actual_output)
-        self.assertEqual(actual_output, expect_output)
-
-    
-    def test_set_api_attr(self):
-        class A():
-            a1 = 1
-        class B():
-            a = A().__class__
-            b1 = 1
-        
-        attr_dict = {"a.a2":2, "b2":2, "b3":3}
-        ApiRegistry.set_api_attr(B, attr_dict)
-
-        for k, v in attr_dict.items():
-            if '.' in k:
-                sub_module_name, sub_op = k.rsplit('.', 1)
-                sub_module = getattr(B, sub_module_name, None)
-                
-                self.assertEqual(getattr(sub_module, sub_op), v)
-            else:
-                self.assertEqual(getattr(B, k), v)
-    
-    def test_api_modularity(self):
-
-        import torch
-        import torch.distributed as dist
-        #import torch_npu   #门禁没有安装torch_npu
-        from msprobe.pytorch.hook_module.api_registry import torch_without_guard_version, npu_distributed_api, is_gpu, torch_version_above_2
-
-        
-
-        reg = ApiRegistry()
-        attr_dict = {"b2":2, "b3":3}
-        reg.tensor_hook_attr = attr_dict
-        reg.torch_hook_attr = attr_dict
-        reg.functional_hook_attr = attr_dict
-        reg.distributed_hook_attr = attr_dict
-        reg.npu_distributed_hook_attr = attr_dict
-        reg.aten_hook_attr = attr_dict
-        reg.vf_hook_attr = attr_dict
-        reg.torch_npu_hook_attr = attr_dict
-
-        reg.api_modularity()
-        self.assertEqual(torch.Tensor.b2, 2)
-
-        self.assertEqual(torch.b2, 2)
-        self.assertEqual(torch.nn.functional.b2, 2)
-        self.assertEqual(dist.b2, 2)
-        self.assertEqual(dist.distributed_c10d.b2, 2)
-        #if not is_gpu and not torch_without_guard_version:
-            #self.assertEqual(torch_npu.distributed.b2, 2)
-            #self.assertEqual(torch_npu.distributed.distributed_c10d.b2, 2)
-        if torch_version_above_2:
-            self.assertEqual(torch.ops.aten.b2, 2)
-        self.assertEqual(torch._VF.b2, 2)
-        #if not is_gpu:
-            #self.assertEqual(torch_npu.b2, 2)
-    
-
-    def test_api_originality(self):
-        import torch
-        import torch.distributed as dist
-        #import torch_npu      #门禁没有安装torch_npu
-        from msprobe.pytorch.hook_module.api_registry import torch_without_guard_version, npu_distributed_api, is_gpu, torch_version_above_2
-
-        
-
-        reg = ApiRegistry()
-        attr_dict = {"b2":2, "b3":3}
-        reg.tensor_hook_attr = attr_dict
-        reg.torch_hook_attr = attr_dict
-        reg.functional_hook_attr = attr_dict
-        reg.distributed_hook_attr = attr_dict
-        reg.npu_distributed_hook_attr = attr_dict
-        reg.aten_hook_attr = attr_dict
-        reg.vf_hook_attr = attr_dict
-        reg.torch_npu_hook_attr = attr_dict
-
-        reg.api_originality()
-        self.assertEqual(torch.Tensor.b2, 2)
-
-        self.assertEqual(torch.b2, 2)
-        self.assertEqual(torch.nn.functional.b2, 2)
-        self.assertEqual(dist.b2, 2)
-        self.assertEqual(dist.distributed_c10d.b2, 2)
-        #if not is_gpu and not torch_without_guard_version:
-            #self.assertEqual(torch_npu.distributed.b2, 2)
-            #self.assertEqual(torch_npu.distributed.distributed_c10d.b2, 2)
-        if torch_version_above_2:
-            self.assertEqual(torch.ops.aten.b2, 2)
-        self.assertEqual(torch._VF.b2, 2)
-        #if not is_gpu:
-            #self.assertEqual(torch_npu.b2, 2)
-
-    def test_initialize_hook(self):
-        def hook_test():
-            pass
-
-        reg = ApiRegistry()
-        reg.initialize_hook(hook_test)
-        empty_list = []
-        self.assertFalse(empty_list==reg.tensor_hook_attr)
-        self.assertFalse(empty_list==reg.torch_hook_attr)
-        self.assertFalse(empty_list==reg.functional_hook_attr)
-        self.assertFalse(empty_list==reg.distributed_hook_attr)
-        self.assertFalse(empty_list==reg.npu_distributed_hook_attr)
-        if torch_version_above_2:
-            #print(True)
-            self.assertFalse(empty_list==reg.aten_hook_attr)
-        if not is_gpu:
-            #print(True)
-            self.assertFalse(empty_list==reg.torch_npu_hook_attr)

msprobe/test/pytorch_ut/hook_module/test_hook_module.py

@@ -1,42 +0,0 @@
-import unittest
-from unittest.mock import patch, Mock
-
-from msprobe.pytorch.hook_module.hook_module import HOOKModule
-
-class TestHookModule(unittest.TestCase):
-    def test_call_1(self):
-        def forward_pre_hook():
-            return "result_input", "result_kwargs"
-        def forward_hook():
-            return 2
-        def backward_hook():
-            pass
-
-        def hook(prefix):
-            return forward_pre_hook, forward_hook, backward_hook
-        HOOKModule.prefix_op_name_ = "123"
-        test = HOOKModule(hook)
-        test._call_func = Mock(return_value=1)
-        result = test()
-        self.assertEqual(result, 1)
-
-    def test_call_2(self):
-        def forward_pre_hook(nope, input, kwargs):
-            return input, kwargs
-        def forward_hook(nope, input, kwargs, result):
-            return input
-        def backward_hook():
-            pass
-
-        def hook(prefix):
-            return forward_pre_hook, forward_hook, backward_hook
-        HOOKModule.prefix_op_name_ = "123"
-        input = 2
-        test = HOOKModule(hook)
-
-        def temp_forward(*input, **kwargs):
-            return input
-
-        test.forward = Mock(return_value=1)
-        result = test(input)
-        self.assertEqual(result, (input, ))

msprobe/test/pytorch_ut/hook_module/test_wrap_aten.py

@@ -1,65 +0,0 @@
-import unittest
-import torch
-from msprobe.pytorch.hook_module.wrap_aten import AtenOPTemplate, AtenOPPacketTemplate
-
-
-def hook(name):
-    def forward_pre_hook(nope, input, kwargs):
-            return input, kwargs
-    def forward_hook(nope, input, kwargs, result):
-            return 2
-    def backward_hook():
-            pass
-    
-    return forward_pre_hook, forward_hook, backward_hook
-
-
-
-class TestWrapAten(unittest.TestCase):
-    def setUp(self):
-        self.aten_op = AtenOPPacketTemplate(torch.ops.aten.convolution, hook)
-        
-    def test_atenop_attribute(self):
-        if torch.__version__.split("+")[0] <= '2.0':
-            return
-        self.setUp()
-        self.assertEqual(self.aten_op.default.op, torch.ops.aten.convolution.default)
-        self.assertEqual(self.aten_op.out.op, torch.ops.aten.convolution.out)
-
-    def test_atenop_forward(self):
-        if torch.__version__.split("+")[0] <= '2.0':
-            return
-        self.setUp()
-        image = torch.randn(4, 3, 24, 24)
-        kernel = torch.randn(10, 3, 3, 3)
-        functional_out = torch.nn.functional.conv2d(image, kernel, stride=[1, 1],
-                                                    padding=[1, 1], dilation=[1, 1], groups=1, bias=None)
-        aten_out = self.aten_op(image, kernel, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 1)
-        self.assertTrue(aten_out == 2)
-
-    def test_atenop_overload_forward(self):
-        if torch.__version__.split("+")[0] <= '2.0':
-            return
-        self.setUp()
-        image = torch.randn(4, 3, 24, 24)
-        kernel = torch.randn(10, 3, 3, 3)
-        functional_out = torch.nn.functional.conv2d(image, kernel, stride=[1, 1],
-                                                    padding=[1, 1], dilation=[1, 1], groups=1, bias=None)
-        aten_out = self.aten_op(image, kernel, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 1)
-        self.assertTrue(aten_out == 2)
-
-    def test_atenop_nonattr(self):
-        if torch.__version__.split("+")[0] <= '2.0':
-            return
-        self.setUp()
-        self.assertRaises(AttributeError, getattr, self.aten_op, "foo")
-
-    def test_atenop_overloads(self):
-        if torch.__version__.split("+")[0] <= '2.0':
-            return
-        self.setUp()
-        self.assertEqual(self.aten_op.overloads(), self.aten_op.opPacket.overloads())
-
-
-
-            

msprobe/test/pytorch_ut/hook_module/test_wrap_distributed.py

@@ -1,35 +0,0 @@
-import unittest
-import torch.distributed as dist
-from msprobe.pytorch.hook_module.wrap_distributed import *
-
-class TestWrapDistributed(unittest.TestCase):
-    def hook(name, prefix):
-        def forward_pre_hook(nope, input, kwargs):
-            return input, kwargs
-        def forward_hook(nope, input, kwargs, result):
-            return 2
-        def backward_hook():
-            pass
-        return forward_pre_hook, forward_hook, backward_hook
-    
-    def test_get_distributed_ops(self):
-        ops = get_distributed_ops()
-        self.assertIsInstance(ops, set)
-
-    def test_DistributedOPTemplate(self):
-        self.setUp()
-        op_name = 'all_reduce'
-        if op_name in get_distributed_ops():
-            op = DistributedOPTemplate(op_name, self.hook)
-            self.assertEqual(op.op_name_, op_name)
-
-    def test_wrap_distributed_op(self):
-        op_name = 'all_reduce'
-        if op_name in get_distributed_ops():
-            wrapped_op = wrap_distributed_op(op_name, self.hook)
-            self.assertTrue(callable(wrapped_op))
-
-    def test_wrap_distributed_ops_and_bind(self):
-        wrap_distributed_ops_and_bind(self.hook)
-        for op_name in get_distributed_ops():
-            self.assertTrue(hasattr(HOOKDistributedOP, "wrap_" + str(op_name)))

msprobe/test/pytorch_ut/hook_module/test_wrap_functional.py

@@ -1,20 +0,0 @@
-import unittest
-import torch
-from msprobe.pytorch.hook_module import wrap_functional as wf
-
-class TestWrapFunctional(unittest.TestCase):
-
-    def test_remove_dropout(self):
-        input_tensor = torch.randn(20, 16)
-        wf.remove_dropout()
-        output_tensor = torch.nn.functional.dropout(input_tensor)
-        self.assertTrue(torch.equal(input_tensor, output_tensor))
-
-    def test_get_functional_ops(self):
-        expected_ops = {'relu', 'sigmoid', 'softmax'}
-        actual_ops = wf.get_functional_ops()
-        self.assertTrue(expected_ops.issubset(actual_ops))
-
-    def test_wrap_functional_ops_and_bind(self):
-        wf.wrap_functional_ops_and_bind(None)
-        self.assertTrue(hasattr(wf.HOOKFunctionalOP, 'wrap_relu'))

msprobe/test/pytorch_ut/hook_module/test_wrap_tensor.py

@@ -1,35 +0,0 @@
-import unittest
-import torch
-import yaml
-from msprobe.pytorch.hook_module.wrap_tensor import get_tensor_ops, HOOKTensor, TensorOPTemplate, wrap_tensor_op, wrap_tensor_ops_and_bind
-
-class TestWrapTensor(unittest.TestCase):
-
-    def hook(name, prefix):
-        def forward_pre_hook(nope, input, kwargs):
-            return input, kwargs
-        def forward_hook(nope, input, kwargs, result):
-            return 2
-        def backward_hook():
-            pass
-        return forward_pre_hook, forward_hook, backward_hook
-    
-    def test_get_tensor_ops(self):
-        result = get_tensor_ops()
-        self.assertIsInstance(result, set)
-
-    def test_HOOKTensor(self):
-        hook_tensor = HOOKTensor()
-        self.assertIsInstance(hook_tensor, HOOKTensor)
-
-    def test_TensorOPTemplate(self):
-        tensor_op_template = TensorOPTemplate('add', self.hook)
-        self.assertTrue(tensor_op_template.op_name_, 'add')
-
-    def test_wrap_tensor_op(self):
-        wrapped_op = wrap_tensor_op('add', self.hook)
-        self.assertTrue(callable(wrapped_op))
-
-    def test_wrap_tensor_ops_and_bind(self):
-        wrap_tensor_ops_and_bind(self.hook)
-        self.assertTrue(hasattr(HOOKTensor, 'wrap_add'))

msprobe/test/pytorch_ut/hook_module/test_wrap_torch.py

@@ -1,43 +0,0 @@
-import unittest
-import torch
-import yaml
-from msprobe.pytorch.hook_module.wrap_torch import *
-
-class TestWrapTorch(unittest.TestCase):
-
-    def hook(name, prefix):
-        def forward_pre_hook(nope, input, kwargs):
-            return input, kwargs
-        def forward_hook(nope, input, kwargs, result):
-            return 2
-        def backward_hook():
-            pass
-        return forward_pre_hook, forward_hook, backward_hook
-    
-    def setUp(self):
-
-        self.op_name = 'add'
-        self.torch_op = wrap_torch_op(self.op_name, self.hook)
-
-    def test_get_torch_ops(self):
-        self.setUp()
-        ops = get_torch_ops()
-        self.assertIsInstance(ops, set)
-        self.assertIn(self.op_name, ops)
-
-    def test_TorchOPTemplate(self):
-        self.setUp()
-        template = TorchOPTemplate(self.op_name, self.hook)
-        self.assertEqual(template.op_name_, self.op_name)        
-        self.assertEqual(template.prefix_op_name_, "Torch." + str(self.op_name) + ".")
-
-    def test_forward(self):
-        self.setUp()
-        template = TorchOPTemplate(self.op_name, self.hook)
-        result = template.forward(torch.tensor([1, 2, 3]), torch.tensor([4, 5, 6]))
-        torch.testing.assert_close(result, torch.tensor([5, 7, 9]))
-
-    def test_wrap_torch_ops_and_bind(self):
-        self.setUp()
-        wrap_torch_ops_and_bind(self.hook)
-        self.assertTrue(hasattr(HOOKTorchOP, "wrap_" + self.op_name))

msprobe/test/pytorch_ut/hook_module/test_wrap_vf.py

@@ -1,11 +0,0 @@
-import unittest
-import torch
-from msprobe.pytorch.hook_module import wrap_vf
-
-class TestWrapVF(unittest.TestCase):
-    def setUp(self):
-        self.hook = lambda x: x
-
-    def test_get_vf_ops(self):
-        ops = wrap_vf.get_vf_ops()
-        self.assertIsInstance(ops, list)

msprobe/test/pytorch_ut/test_pt_config.py

@@ -1,69 +0,0 @@
-from unittest import TestCase
-from unittest.mock import patch, mock_open
-
-from msprobe.core.common.const import Const
-from msprobe.pytorch.pt_config import parse_json_config, parse_task_config
-
-
-class TestPtConfig(TestCase):
-    def test_parse_json_config(self):
-        mock_json_data = {
-            "task": "statistics",
-            "dump_path": "./dump/",
-            "rank": [],
-            "step": [],
-            "level": "L1",
-            "seed": 1234,
-            "statistics": {
-                "scope": [],
-                "list": [],
-                "data_mode": ["all"],
-            },
-            "tensor": {
-                "file_format": "npy"
-            }
-        }
-        with patch("msprobe.pytorch.pt_config.os.path.join", return_value="/path/config.json"), \
-                patch("msprobe.pytorch.pt_config.FileOpen", mock_open(read_data='')), \
-                patch("msprobe.pytorch.pt_config.json.load", return_value=mock_json_data):
-            common_config, task_config = parse_json_config(None, None)
-        self.assertEqual(common_config.task, Const.STATISTICS)
-        self.assertEqual(task_config.data_mode, ["all"])
-
-        with patch("msprobe.pytorch.pt_config.os.path.join", return_value="/path/config.json"), \
-                patch("msprobe.pytorch.pt_config.FileOpen", mock_open(read_data='')), \
-                patch("msprobe.pytorch.pt_config.json.load", return_value=mock_json_data):
-            common_config, task_config = parse_json_config(None, Const.TENSOR)
-        self.assertEqual(common_config.task, Const.STATISTICS)
-        self.assertEqual(task_config.file_format, "npy")
-
-    def test_parse_task_config(self):
-        overflow_check_config = {
-            "overflow_check": {
-                "overflow_nums": 1,
-                "check_mode": "all"
-            }
-        }
-        result = parse_task_config(Const.OVERFLOW_CHECK, overflow_check_config)
-        self.assertEqual(result.overflow_num, 1)
-        self.assertEqual(result.check_mode, "all")
-
-        free_benchmark_config = {
-            "free_benchmark": {
-                "scope": [],
-                "list": ["conv2d"],
-                "fuzz_device": "npu",
-                "pert_mode": "improve_precision",
-                "handler_type": "check",
-                "fuzz_level": "L1",
-                "fuzz_stage": "forward",
-                "if_preheat": False,
-                "preheat_step": 15,
-                "max_sample": 20
-            }
-        }
-        result = parse_task_config(Const.FREE_BENCHMARK, free_benchmark_config)
-        self.assertEqual(result.pert_mode, "improve_precision")
-        self.assertEqual(result.handler_type, "check")
-        self.assertEqual(result.preheat_step, 15)
-        self.assertEqual(result.max_sample, 20)

msprobe/test/pytorch_ut/test_service.py

@@ -1,59 +0,0 @@
-import unittest
-from unittest.mock import patch, mock_open
-
-import torch.nn as nn
-from msprobe.core.common.utils import Const
-from msprobe.pytorch.debugger.debugger_config import DebuggerConfig
-from msprobe.pytorch.pt_config import parse_json_config
-from msprobe.pytorch.service import Service
-
-
-class TestService(unittest.TestCase):
-    def setUp(self):
-        mock_json_data = {
-            "dump_path": "./dump/",
-        }
-        with patch("msprobe.pytorch.pt_config.FileOpen", mock_open(read_data='')), \
-                patch("msprobe.pytorch.pt_config.json.load", return_value=mock_json_data):
-            common_config, task_config = parse_json_config("./config.json", Const.STATISTICS)
-        self.config = DebuggerConfig(common_config, task_config, Const.STATISTICS, "./ut_dump", "L1")
-        self.service = Service(self.config)
-
-    def test_start(self):
-        with patch("msprobe.pytorch.service.get_rank_if_initialized", return_value=0), \
-                patch("msprobe.pytorch.service.Service.create_dirs", return_value=None):
-            self.service.start(None)
-        self.assertEqual(self.service.current_rank, 0)
-
-    def test_stop_and_step(self):
-        with patch("msprobe.core.data_dump.data_collector.DataCollector.write_json", return_value=None):
-            self.service.stop()
-        self.assertFalse(self.service.switch)
-
-        self.service.step()
-        self.assertEqual(self.service.current_iter, 1)
-
-    def test_register_hook_new(self):
-        class TestModule(nn.Module):
-            def __init__(self) -> None:
-                super().__init__()
-                self.linear = nn.Linear(in_features=8, out_features=4)
-
-            def forward(self, x):
-                x = self.linear(x)
-                return x
-
-        self.service.model = TestModule()
-        self.config.level = "L0"
-        with patch("msprobe.pytorch.service.logger.info_on_rank_0") as mock_logger, \
-                patch("msprobe.pytorch.service.remove_dropout", return_value=None):
-            self.service.register_hook_new()
-            self.assertEqual(mock_logger.call_count, 2)
-
-    def test_create_dirs(self):
-        with patch("msprobe.pytorch.service.Path.mkdir", return_value=None), \
-                patch("msprobe.core.common.file_check.FileChecker.common_check", return_value=None), \
-                patch("msprobe.core.data_dump.data_collector.DataCollector.update_dump_paths",
-                      return_value=None):
-            self.service.create_dirs()
-        self.assertEqual(self.service.dump_iter_dir, "./ut_dump/step0")

msprobe/test/resources/advisor.txt

@@ -1,3 +0,0 @@
-Line: NA
-Suspect Nodes: NA
-Expert Advice: All data in comparison result meets the accuracy requirements.

msprobe/test/resources/compare_result_20230703104808.csv

@@ -1,9 +0,0 @@
-NPU Name,Bench Name,NPU Tensor Dtype,Bench Tensor Dtype,NPU Tensor Shape,Bench Tensor Shape,Cosine,MaxAbsErr,NPU max,NPU min,NPU mean,Bench max,Bench min,Bench mean,Accuracy Reached or Not,Err_message
-Functional_linear_0_forward_input.0,Functional_linear_0_forward_input.0,torch.float32,torch.float32,"[3, 2]","[3, 2]",1.0,0.000000,1.948258399963379,-1.0052297115325928,-0.2003595232963562,1.948258399963379,-1.0052297115325928,-0.2003595232963562,Yes,
-Functional_linear_0_forward_input.1,Functional_linear_0_forward_input.1,torch.float32,torch.float32,"[3, 2]","[3, 2]",1.0,0.000000,0.28375449776649475,-0.6661239266395569,-0.2789986729621887,0.28375449776649475,-0.6661239266395569,-0.2789986729621887,Yes,
-Functional_linear_0_forward_input.2,Functional_linear_0_forward_input.2,torch.float32,torch.float32,[3],[3],1.0,0.000000,0.2457989901304245,-0.6338542103767395,-0.14437106251716614,0.2457989901304245,-0.6338542103767395,-0.14437106251716614,Yes,
-Functional_linear_0_forward_output,Functional_linear_0_forward_output,torch.float32,torch.float32,"[3, 3]","[3, 3]",1.0,0.000000,0.8278868794441223,-0.8729169964790344,0.16790540516376495,0.8278868794441223,-0.8729169964790344,0.16790540516376495,Yes,
-Torch_relu_0_forward_input.0,Torch_relu_0_forward_input.0,torch.float32,torch.float32,"[3, 3]","[3, 3]",1.0,0.000000,0.8278868794441223,-0.8729169964790344,0.16790540516376495,0.8278868794441223,-0.8729169964790344,0.16790540516376495,Yes,
-Torch_relu_0_forward_output,Torch_relu_0_forward_output,torch.float32,torch.float32,"[3, 3]","[3, 3]",1.0,0.000000,0.8278868794441223,0.0,0.31367552280426025,0.8278868794441223,0.0,0.31367552280426025,Yes,
-Functional_relu_0_forward_input.0,Functional_relu_0_forward_input.0,torch.float32,torch.float32,"[3, 3]","[3, 3]",1.0,0.000000,0.8278868794441223,-0.8729169964790344,0.16790540516376495,0.8278868794441223,-0.8729169964790344,0.16790540516376495,Yes,
-Functional_relu_0_forward_output,Functional_relu_0_forward_output,torch.float32,torch.float32,"[3, 3]","[3, 3]",1.0,0.000000,0.8278868794441223,0.0,0.31367552280426025,0.8278868794441223,0.0,0.31367552280426025,Yes,

msprobe/test/resources/compare_result_without_accuracy.csv

@@ -1,9 +0,0 @@
-NPU Name,Bench Name,NPU Tensor Dtype,Bench Tensor Dtype,NPU Tensor Shape,Bench Tensor Shape,Cosine,MaxAbsErr,NPU max,NPU min,NPU mean,Bench max,Bench min,Bench mean,Accuracy Reached or Not,Err_message
-,Functional_linear_0_forward_input.0,torch.float32,torch.float32,"[3, 2]","[3, 2]",1,0,1.9482584,-1.005229712,-0.200359523,1.9482584,-1.005229712,-0.200359523,,
-,Functional_linear_0_forward_input.1,torch.float32,torch.float32,"[3, 2]","[3, 2]",1,0,0.283754498,-0.666123927,-0.278998673,0.283754498,-0.666123927,-0.278998673,,
-,Functional_linear_0_forward_input.2,torch.float32,torch.float32,[3],[3],1,0,0.24579899,-0.63385421,-0.144371063,0.24579899,-0.63385421,-0.144371063,,
-,Functional_linear_0_forward_output,torch.float32,torch.float32,"[3, 3]","[3, 3]",1,0,0.827886879,-0.872916996,0.167905405,0.827886879,-0.872916996,0.167905405,,
-,Torch_relu_0_forward_input.0,torch.float32,torch.float32,"[3, 3]","[3, 3]",1,0,0.827886879,-0.872916996,0.167905405,0.827886879,-0.872916996,0.167905405,,
-,Torch_relu_0_forward_output,torch.float32,torch.float32,"[3, 3]","[3, 3]",1,0,0.827886879,0,0.313675523,0.827886879,0,0.313675523,,
-,Functional_relu_0_forward_input.0,torch.float32,torch.float32,"[3, 3]","[3, 3]",1,0,0.827886879,-0.872916996,0.167905405,0.827886879,-0.872916996,0.167905405,,
-,Functional_relu_0_forward_output,torch.float32,torch.float32,"[3, 3]","[3, 3]",1,0,0.827886879,0,0.313675523,0.827886879,0,0.313675523,,

msprobe/test/resources/config.yaml

@@ -1,3 +0,0 @@
-white_list: []
-error_data_path: './'
-precision: 14

msprobe/test/resources/npu_test.pkl

@@ -1,8 +0,0 @@
-["Functional_linear_0_forward_input.0", 1, [], "torch.float32", [3, 2], [1.948258399963379, -1.0052297115325928, -0.2003595232963562]]
-["Functional_linear_0_forward_input.1", 1, [], "torch.float32", [3, 2], [0.28375449776649475, -0.6661239266395569, -0.2789986729621887]]
-["Functional_linear_0_forward_input.2", 1, [], "torch.float32", [3], [0.2457989901304245, -0.6338542103767395, -0.14437106251716614]]
-["Functional_linear_0_forward_output", 1, [], "torch.float32", [3, 3], [0.8278868794441223, -0.8729169964790344, 0.16790540516376495]]
-["Torch_relu_0_forward_input.0", 1, [], "torch.float32", [3, 3], [0.8278868794441223, -0.8729169964790344, 0.16790540516376495]]
-["Torch_relu_0_forward_output", 1, [], "torch.float32", [3, 3], [0.8278868794441223, 0.0, 0.31367552280426025]]
-["Functional_relu_0_forward_input.0", 1, [], "torch.float32", [3, 3], [0.8278868794441223, -0.8729169964790344, 0.16790540516376495]]
-["Functional_relu_0_forward_output", 1, [], "torch.float32", [3, 3], [0.8278868794441223, 0.0, 0.31367552280426025]]

msprobe/test/run_test.sh

@@ -1,30 +0,0 @@
-#!/bin/bash
-CUR_DIR=$(dirname $(readlink -f $0))
-TOP_DIR=${CUR_DIR}/..
-TEST_DIR=${TOP_DIR}/"test"
-SRC_DIR=${TOP_DIR}/../
-
-install_pytest() {
-    if ! pip show pytest &> /dev/null; then
-        echo "pytest not found, trying to install..."
-        pip install pytest
-    fi
-
-    if ! pip show pytest-cov &> /dev/null; then
-        echo "pytest-cov not found, trying to install..."
-        pip install pytest-cov
-    fi
-}
-
-run_ut() {
-    install_pytest
-
-    export PYTHONPATH=${SRC_DIR}:${PYTHONPATH}
-    python3 run_ut.py
-}
-
-main() {
-    cd ${TEST_DIR} && run_ut
-}
-
-main $@

msprobe/test/run_ut.py

@@ -1,58 +0,0 @@
-import os
-import shutil
-import subprocess
-import sys
-
-from msprobe.core.common.log import logger
-
-
-def run_ut():
-    cur_dir = os.path.realpath(os.path.dirname(__file__))
-    ut_path = cur_dir
-    cov_dir = os.path.dirname(cur_dir)
-    report_dir = os.path.join(cur_dir, "report")
-    final_xml_path = os.path.join(report_dir, "final.xml")
-    cov_report_path = os.path.join(report_dir, "coverage.xml")
-
-    if os.path.exists(report_dir):
-        shutil.rmtree(report_dir)
-    os.makedirs(report_dir)
-
-    pytest_cmd = [
-                     "python3", "-m", "pytest",
-                     ut_path,
-                     f"--junitxml={final_xml_path}",
-                     f"--cov={cov_dir}",
-                     "--cov-branch",
-                     f"--cov-report=xml:{cov_report_path}",
-                 ]
-
-    try:
-        with subprocess.Popen(
-                pytest_cmd,
-                shell=False,
-                stdout=subprocess.PIPE,
-                stderr=subprocess.STDOUT,
-                text=True,
-        ) as proc:
-            for line in proc.stdout:
-                logger.info(line.strip())
-
-            proc.wait()
-
-            if proc.returncode == 0:
-                logger.info("Unit tests executed successfully.")
-                return True
-            else:
-                logger.error("Unit tests execution failed.")
-                return False
-    except Exception as e:
-        logger.error(f"An error occurred during test execution: {e}")
-        return False
-
-
-if __name__ == "__main__":
-    if run_ut():
-        sys.exit(0)
-    else:
-        sys.exit(1)