Unlock 30% off on Manual Testing Annual Plans this Holiday Season.

Copied to Clipboard!

How to use reset_model method in Behave

Best Python code snippet using behave

test_base.py

Source:test_base.py

...58        pt_outputs = [o.detach().numpy() for o in pt_outputs]59      for l, r in zip(ort_outputs, pt_outputs):60        assert np.allclose(l, r, atol=1e-4, rtol=1e-4, equal_nan=True)61  def test_conv_flatten_relu(self):62    reset_model(13)63    nps = [np.random.randn(1, 3, 224, 224).astype(np.float32)]64    inputs = Input(*nps)65    conv_node = Conv(inputs[0],66                     np.random.randn(32, 3, 3, 3).astype(np.float32),67                     np.random.randn(32).astype(np.float32))68    flatten_node = Flatten(conv_node)69    relu_node = Relu(flatten_node)70    Output(relu_node)71    self._run(list(zip(inputs, nps)))72  def test_conv_batchnorm_maxpool_flatten_add_relu(self):73    reset_model(13)74    nps = [np.random.randn(1, 3, 224, 224).astype(np.float32)]75    inputs = Input(*nps)76    conv_node = Conv(inputs[0],77                     np.random.randn(32, 3, 3, 3).astype(np.float32),78                     np.random.randn(32).astype(np.float32))79    bn_node = BatchNormalization(80        conv_node,81        np.ones(32,).astype(np.float32),82        np.zeros(32,).astype(np.float32),83        np.random.randn(32).astype(np.float32),84        np.abs(np.random.randn(32).astype(np.float32)),85    )86    max_pool_node = MaxPool(bn_node,87                            kernel_shape=(3, 3),88                            strides=(2, 2),89                            pads=(0, 0, 1, 1))90    flatten_node = Flatten(max_pool_node, axis=1)91    add_node = Add(flatten_node, np.random.randn(1).astype(np.float32))92    relu_node = Relu(add_node)93    Output(relu_node)94    self._run(list(zip(inputs, nps)))95  def test_abs(self):96    reset_model(13)97    nps = [np.random.randn(1, 10).astype(np.float32)]98    inputs = Input(*nps)99    Output(Abs(inputs[0]))100    self._run(list(zip(inputs, nps)))101  def test_acos(self):102    reset_model(13)103    nps = [np.random.randn(5).astype(np.float32)]104    inputs = Input(*nps)105    Output(Acos(inputs[0]))106    self._run(list(zip(inputs, nps)))107  def test_acosh(self):108    reset_model(13)109    nps = [np.random.randn(5).astype(np.float32)]110    inputs = Input(*nps)111    Output(Acosh(inputs[0]))112    self._run(list(zip(inputs, nps)))113  def test_tanh(self):114    reset_model(13)115    nps = [np.random.randn(5).astype(np.float32)]116    inputs = Input(*nps)117    Output(Tanh(inputs[0]))118    self._run(list(zip(inputs, nps)))119  def test_add(self):120    reset_model(13)121    nps = [np.random.randn(1, 10).astype(np.float32)]122    inputs = Input(*nps)123    Output(Add(inputs[0], np.random.randn(1, 10).astype(np.float32)))124    self._run(list(zip(inputs, nps)))125  def test_sub(self):126    reset_model(13)127    nps = [np.random.randn(1, 10).astype(np.float32)]128    inputs = Input(*nps)129    Output(Sub(inputs[0], np.random.randn(1, 10).astype(np.float32)))130    self._run(list(zip(inputs, nps)))131  def test_and(self):132    reset_model(13)133    nps = [134        np.random.randint(low=0, high=1, size=(5,)).astype(bool),135        np.random.randint(low=0, high=1, size=(5,)).astype(bool)136    ]137    inputs = Input(*nps)138    Output(And(*inputs))139    self._run(list(zip(inputs, nps)))140  def test_argmax(self):141    reset_model(13)142    nps = [np.random.randn(1, 10).astype(np.float32)]143    inputs = Input(*nps)144    Output(ArgMax(inputs, axis=1))145    self._run(list(zip(inputs, nps)))146  def test_argmin(self):147    reset_model(13)148    nps = [np.random.randn(1, 10).astype(np.float32)]149    inputs = Input(*nps)150    Output(ArgMin(inputs, axis=1))151    self._run(list(zip(inputs, nps)))152  def test_asin(self):153    reset_model(13)154    nps = [np.random.randn(5).astype(np.float32)]155    inputs = Input(*nps)156    Output(Asin(inputs[0]))157    self._run(list(zip(inputs, nps)))158  def test_asinh(self):159    reset_model(13)160    nps = [np.random.randn(5).astype(np.float32)]161    inputs = Input(*nps)162    Output(Asinh(inputs[0]))163    self._run(list(zip(inputs, nps)))164  def test_atan(self):165    reset_model(13)166    nps = [np.random.randn(5).astype(np.float32)]167    inputs = Input(*nps)168    Output(Atan(inputs[0]))169    self._run(list(zip(inputs, nps)))170  def test_atanh(self):171    reset_model(13)172    nps = [np.random.randn(5).astype(np.float32)]173    inputs = Input(*nps)174    Output(Atanh(inputs[0]))175    self._run(list(zip(inputs, nps)))176  def test_avg_pool(self):177    reset_model(13)178    nps = [np.random.randn(1, 1, 5, 5).astype(np.float32)]179    inputs = Input(*nps)180    Output(AveragePool(inputs, kernel_shape=(3, 3), pads=(0, 0, 1, 1)))181    self._run(list(zip(inputs, nps)))182  def test_avg_pool_no_pad(self):183    reset_model(13)184    nps = [np.random.randn(1, 1, 6, 6).astype(np.float32)]185    inputs = Input(*nps)186    Output(AveragePool(inputs, kernel_shape=(3, 3)))187    self._run(list(zip(inputs, nps)))188  def test_bitshift_left(self):189    reset_model(13)190    nps = [np.array([1, 2]).astype(np.uint8), np.array([1, 2]).astype(np.uint8)]191    inputs = Input(*nps)192    Output(BitShift(*inputs, direction="LEFT"))193    self._run(list(zip(inputs, nps)))194  def test_bitshift_right(self):195    reset_model(13)196    nps = [np.array([1, 4]).astype(np.uint8), np.array([1, 1]).astype(np.uint8)]197    inputs = Input(*nps)198    Output(BitShift(*inputs, direction="RIGHT"))199    self._run(list(zip(inputs, nps)))200  def test_batch_normalization(self):201    reset_model(13)202    nps = [np.random.randn(1, 32, 3, 3).astype(np.float32)]203    inputs = Input(*nps)204    Output(BatchNormalization(205        inputs[0],206        np.ones(32,).astype(np.float32),207        np.zeros(32,).astype(np.float32),208        np.random.randn(32).astype(np.float32),209        np.abs(np.random.randn(32).astype(np.float32)),210    ),211           output_num=1)212    self._run(list(zip(inputs, nps)))213  def test_cast(self):214    reset_model(13)215    nps = [216        np.random.randn(1, 10).astype(np.float32),217    ]218    inputs = Input(*nps)219    Output(Cast(inputs, to=6))220    self._run(list(zip(inputs, nps)))221  def test_ceil(self):222    reset_model(13)223    nps = [np.random.randn(1, 10).astype(np.float32)]224    inputs = Input(*nps)225    Output(Ceil(inputs[0]))226    self._run(list(zip(inputs, nps)))227  def test_clip(self):228    reset_model(13)229    nps = [230        np.random.randn(1, 5).astype(np.float32),231        np.asarray(0).astype(np.float32)232    ]233    inputs = Input(*nps)234    Output(Clip(*inputs))235    self._run(list(zip(inputs, nps)))236  def test_concat(self):237    reset_model(13)238    nps = [239        np.random.randn(1, 10).astype(np.float32),240        np.random.randn(2, 10).astype(np.float32)241    ]242    inputs = Input(*nps)243    Output(Concat(inputs, axis=0))244    self._run(list(zip(inputs, nps)))245  def test_constant_add(self):246    reset_model(13)247    nps = [248        np.random.randn(1, 10).astype(np.float32),249    ]250    inputs = Input(*nps)251    constant_value = np.random.randn(1, 10).astype(np.float32)252    t = make_tensor("", 1, constant_value.shape, constant_value.flatten())253    Output(Add(inputs, Constant(value=t)))254    self._run(list(zip(inputs, nps)))255  def test_constant_of_shape(self):256    reset_model(13)257    nps = [258        np.array([2, 3]).astype(np.int64),259    ]260    inputs = Input(*nps)261    constant_value = np.random.randn(1).astype(np.float32)262    t = make_tensor("", 1, constant_value.shape, constant_value)263    Output(ConstantOfShape(inputs, value=t))264    self._run(list(zip(inputs, nps)))265  def test_conv(self):266    reset_model(13)267    nps = [np.random.randn(1, 3, 3, 3).astype(np.float32)]268    inputs = Input(*nps)269    Output(270        Conv(271            inputs[0],272            np.random.randn(32, 3, 3, 3).astype(np.float32),273            np.random.randn(32).astype(np.float32),274        ))275    self._run(list(zip(inputs, nps)))276  def test_conv_with_autopad_same(self):277    reset_model(13)278    x = np.array([[[279        [0., 1., 2., 3., 4.],  # (1, 1, 5, 5) input tensor280        [5., 6., 7., 8., 9.],281        [10., 11., 12., 13., 14.],282        [15., 16., 17., 18., 19.],283        [20., 21., 22., 23., 24.]284    ]]]).astype(np.float32)285    W = np.array([[[286        [1., 1., 1.],  # (1, 1, 3, 3) tensor for convolution weights287        [1., 1., 1.],288        [1., 1., 1.]289    ]]]).astype(np.float32)290    nps = [x]291    inputs = Input(*nps)292    Output(293        Conv(294            inputs[0],295            W,296            auto_pad='SAME_LOWER',297            kernel_shape=[3, 3],298            strides=[2, 2],299        ))300    self._run(list(zip(inputs, nps)))301  def test_conv_transpose(self):302    reset_model(13)303    nps = [np.random.randn(1, 3, 3, 3).astype(np.float32)]304    inputs = Input(*nps)305    Output(306        ConvTranspose(307            inputs[0],308            np.random.randn(3, 32, 3, 3).astype(np.float32),309            np.random.randn(32).astype(np.float32),310        ))311    self._run(list(zip(inputs, nps)))312  def test_conv_transpose_pads(self):313    reset_model(13)314    nps = [np.random.randn(1, 1, 3, 3).astype(np.float32)]315    inputs = Input(*nps)316    Output(317        ConvTranspose(inputs[0],318                      np.random.randn(1, 2, 3, 3).astype(np.float32),319                      strides=[3, 2],320                      pads=[1, 2, 1, 2]))321    self._run(list(zip(inputs, nps)))322  def test_conv_transpose_dilations(self):323    reset_model(13)324    nps = [np.random.randn(1, 1, 3, 3).astype(np.float32)]325    inputs = Input(*nps)326    Output(327        ConvTranspose(328            inputs[0],329            np.random.randn(1, 1, 2, 2).astype(np.float32),330            dilations=[2, 2],331        ))332    self._run(list(zip(inputs, nps)))333  def test_conv_transpose_attributes(self):334    reset_model(13)335    nps = [np.random.randn(1, 1, 3, 3).astype(np.float32)]336    inputs = Input(*nps)337    Output(338        ConvTranspose(339            inputs[0],340            np.random.randn(1, 2, 3, 3).astype(np.float32),341            strides=[3, 2],342            output_shape=[10, 8],343            output_padding=[1, 1],344        ))345    self._run(list(zip(inputs, nps)))346  def test_cos(self):347    reset_model(13)348    nps = [np.random.randn(5).astype(np.float32)]349    inputs = Input(*nps)350    Output(Cos(inputs[0]))351    self._run(list(zip(inputs, nps)))352  def test_cosh(self):353    reset_model(13)354    nps = [np.random.randn(5).astype(np.float32)]355    inputs = Input(*nps)356    Output(Cosh(inputs[0]))357    self._run(list(zip(inputs, nps)))358  def test_flatten(self):359    reset_model(13)360    nps = [np.random.randn(1, 3, 3, 3).astype(np.float32)]361    inputs = Input(*nps)362    Output(Flatten(inputs))363    self._run(list(zip(inputs, nps)))364  def test_gather(self):365    reset_model(13)366    nps = [367        np.random.randn(6, 8, 3).astype(np.float32),368        np.random.randn(*[3, 2]).astype(np.int64),369    ]370    inputs = Input(*nps)371    Output(Gather(*inputs, axis=0))372    self._run(list(zip(inputs, nps)))373  def test_gather_axis_1(self):374    reset_model(13)375    nps = [376        np.random.randn(6, 3, 3).astype(np.float32),377        np.array([[0, 2], [0, 1], [2, 0]]).astype(np.int64),378    ]379    inputs = Input(*nps)380    Output(Gather(*inputs, axis=1))381    self._run(list(zip(inputs, nps)))382  def test_gather_nd(self):383    reset_model(13)384    nps = [385        np.random.randn(6, 8, 3).astype(np.float32),386        np.random.randn(*[3, 2]).astype(np.int64),387    ]388    inputs = Input(*nps)389    Output(GatherND(*inputs, batch_dims=0))390    self._run(list(zip(inputs, nps)))391  @pytest.mark.skip(reason="Not implemented for batch_dims != 0")392  def test_gather_nd_batch_dims_1(self):393    reset_model(13)394    nps = [395        np.array([[[0, 1], [2, 3]], [[4, 5], [6, 7]]]),396        np.array([[1], [0]]).astype(np.int64),397    ]398    inputs = Input(*nps)399    Output(GatherND(*inputs, batch_dims=1))400    self._run(list(zip(inputs, nps)))401  def test_gemm(self):402    reset_model(13)403    nps = [404        np.random.randn(2, 3).astype(np.float32),405        np.random.randn(3, 4).astype(np.float32),406        np.random.randn(2, 4).astype(np.float32),407    ]408    inputs = Input(*nps)409    Output(Gemm(*inputs))410    self._run(list(zip(inputs, nps)))411  def test_global_average_pool(self):412    reset_model(13)413    nps = [np.random.randn(2, 3, 4).astype(np.float32)]414    inputs = Input(*nps)415    Output(GlobalAveragePool(*inputs))416    self._run(list(zip(inputs, nps)))417  def test_mat_mul(self):418    reset_model(13)419    nps = [420        np.random.randn(5, 2, 3).astype(np.float32),421        np.random.randn(5, 3, 2).astype(np.float32)422    ]423    inputs = Input(*nps)424    Output(MatMul(*inputs))425    self._run(list(zip(inputs, nps)))426  def test_max(self):427    reset_model(13)428    nps = [429        np.random.randn(1, 10).astype(np.float32),430        np.random.randn(2, 10).astype(np.float32)431    ]432    inputs = Input(*nps)433    Output(Max(inputs))434    self._run(list(zip(inputs, nps)))435  def test_max_pool(self):436    reset_model(13)437    nps = [np.random.randn(1, 1, 5, 5).astype(np.float32)]438    inputs = Input(*nps)439    Output(MaxPool(inputs, kernel_shape=(3, 3), pads=(0, 0, 1, 1)),440           output_num=1)441    self._run(list(zip(inputs, nps)))442  def test_max_pool_pads(self):443    reset_model(13)444    nps = [np.random.randn(1, 1, 4, 4).astype(np.float32)]445    inputs = Input(*nps)446    Output(MaxPool(inputs, kernel_shape=(3, 3), pads=(1, 1, 1, 1)),447           output_num=1)448    self._run(list(zip(inputs, nps)))449  def test_mul(self):450    reset_model(13)451    nps = [452        np.random.randn(1, 2, 3).astype(np.float32),453        np.random.randn(1, 2, 3).astype(np.float32)454    ]455    inputs = Input(*nps)456    Output(Mul(*inputs))457    self._run(list(zip(inputs, nps)))458  def test_non_max_suppression_center_point_box_format(self):459    reset_model(13)460    boxes = np.array([[[0.5, 0.5, 1.0, 1.0], [0.5, 0.6, 1.0, 1.0],461                       [0.5, 0.4, 1.0, 1.0], [0.5, 10.5, 1.0, 1.0],462                       [0.5, 10.6, 1.0, 1.0], [0.5, 100.5, 1.0,463                                               1.0]]]).astype(np.float32)464    scores = np.array([[[0.9, 0.75, 0.6, 0.95, 0.5, 0.3]]]).astype(np.float32)465    max_output_boxes_per_class = np.array([3]).astype(np.int64)466    iou_threshold = np.array([0.5]).astype(np.float32)467    score_threshold = np.array([0.0]).astype(np.float32)468    selected_indices = np.array([[0, 0, 3], [0, 0, 0], [0, 0,469                                                        5]]).astype(np.int64)470    inputs = Input(*[boxes, scores])471    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,472                             iou_threshold))473    self._run(list(zip(inputs, [boxes, scores])))474  def test_non_max_suppression_flipped_coordinates(self):475    reset_model(13)476    boxes = np.array([[[1.0, 1.0, 0.0, 0.0], [0.0, 0.1, 1.0, 1.1],477                       [0.0, 0.9, 1.0, -0.1], [0.0, 10.0, 1.0, 11.0],478                       [1.0, 10.1, 0.0, 11.1], [1.0, 101.0, 0.0,479                                                100.0]]]).astype(np.float32)480    scores = np.array([[[0.9, 0.75, 0.6, 0.95, 0.5, 0.3]]]).astype(np.float32)481    max_output_boxes_per_class = np.array([5]).astype(np.int64)482    iou_threshold = np.array([0.5]).astype(np.float32)483    score_threshold = np.array([0.0]).astype(np.float32)484    selected_indices = np.array([[0, 0, 3], [0, 0, 0], [0, 0,485                                                        5]]).astype(np.int64)486    inputs = Input(*[boxes, scores])487    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,488                             iou_threshold))489    self._run(list(zip(inputs, [boxes, scores])))490  def test_non_max_suppression_identical_boxes(self):491    reset_model(13)492    boxes = np.array([[[0.0, 0.0, 1.0, 1.0], [0.0, 0.0, 1.0, 1.0],493                       [0.0, 0.0, 1.0, 1.0], [0.0, 0.0, 1.0, 1.0],494                       [0.0, 0.0, 1.0, 1.0], [0.0, 0.0, 1.0, 1.0],495                       [0.0, 0.0, 1.0, 1.0], [0.0, 0.0, 1.0, 1.0],496                       [0.0, 0.0, 1.0, 1.0], [0.0, 0.0, 1.0,497                                              1.0]]]).astype(np.float32)498    scores = np.array([[[0.9, 0.9, 0.9, 0.9, 0.9, 0.9, 0.9, 0.9, 0.9,499                         0.9]]]).astype(np.float32)500    max_output_boxes_per_class = np.array([3]).astype(np.int64)501    iou_threshold = np.array([0.5]).astype(np.float32)502    score_threshold = np.array([0.0]).astype(np.float32)503    selected_indices = np.array([[0, 0, 0]]).astype(np.int64)504    inputs = Input(*[boxes, scores])505    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,506                             iou_threshold))507    self._run(list(zip(inputs, [boxes, scores])))508  def test_non_max_suppression_limit_output_size(self):509    reset_model(13)510    boxes = np.array([[[0.0, 0.0, 1.0, 1.0], [0.0, 0.1, 1.0, 1.1],511                       [0.0, -0.1, 1.0, 0.9], [0.0, 10.0, 1.0, 11.0],512                       [0.0, 10.1, 1.0, 11.1], [0.0, 100.0, 1.0,513                                                101.0]]]).astype(np.float32)514    scores = np.array([[[0.9, 0.75, 0.6, 0.95, 0.5, 0.3]]]).astype(np.float32)515    max_output_boxes_per_class = np.array([2]).astype(np.int64)516    iou_threshold = np.array([0.5]).astype(np.float32)517    score_threshold = np.array([0.0]).astype(np.float32)518    selected_indices = np.array([[0, 0, 3], [0, 0, 0]]).astype(np.int64)519    inputs = Input(*[boxes, scores])520    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,521                             iou_threshold))522    self._run(list(zip(inputs, [boxes, scores])))523  def test_non_max_suppression_single_box(self):524    reset_model(13)525    boxes = np.array([[[0.0, 0.0, 1.0, 1.0]]]).astype(np.float32)526    scores = np.array([[[0.9]]]).astype(np.float32)527    max_output_boxes_per_class = np.array([3]).astype(np.int64)528    iou_threshold = np.array([0.5]).astype(np.float32)529    score_threshold = np.array([0.0]).astype(np.float32)530    inputs = Input(*[boxes, scores])531    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,532                             iou_threshold))533    self._run(list(zip(inputs, [boxes, scores])))534  def test_non_max_suppression_suppress_by_IOU(self):535    reset_model(13)536    boxes = np.array([[[0.0, 0.0, 1.0, 1.0], [0.0, 0.1, 1.0, 1.1],537                       [0.0, -0.1, 1.0, 0.9], [0.0, 10.0, 1.0, 11.0],538                       [0.0, 10.1, 1.0, 11.1], [0.0, 100.0, 1.0,539                                                101.0]]]).astype(np.float32)540    scores = np.array([[[0.9, 0.75, 0.6, 0.95, 0.5, 0.3]]]).astype(np.float32)541    max_output_boxes_per_class = np.array([3]).astype(np.int64)542    iou_threshold = np.array([0.5]).astype(np.float32)543    score_threshold = np.array([0.0]).astype(np.float32)544    selected_indices = np.array([[0, 0, 3], [0, 0, 0], [0, 0,545                                                        5]]).astype(np.int64)546    inputs = Input(*[boxes, scores])547    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,548                             iou_threshold))549    self._run(list(zip(inputs, [boxes, scores])))550  def test_non_max_suppression_suppress_by_IOU_and_scores(self):551    reset_model(13)552    boxes = np.array([[[0.0, 0.0, 1.0, 1.0], [0.0, 0.1, 1.0, 1.1],553                       [0.0, -0.1, 1.0, 0.9], [0.0, 10.0, 1.0, 11.0],554                       [0.0, 10.1, 1.0, 11.1], [0.0, 100.0, 1.0,555                                                101.0]]]).astype(np.float32)556    scores = np.array([[[0.9, 0.75, 0.6, 0.95, 0.5, 0.3]]]).astype(np.float32)557    max_output_boxes_per_class = np.array([3]).astype(np.int64)558    iou_threshold = np.array([0.5]).astype(np.float32)559    score_threshold = np.array([0.4]).astype(np.float32)560    selected_indices = np.array([[0, 0, 3], [0, 0, 0]]).astype(np.int64)561    inputs = Input(*[boxes, scores])562    Output(563        NonMaxSuppression(*inputs, max_output_boxes_per_class, iou_threshold,564                          score_threshold))565    self._run(list(zip(inputs, [boxes, scores])))566  def test_non_max_suppression_two_batches(self):567    reset_model(13)568    boxes = np.array([[[0.0, 0.0, 1.0, 1.0], [0.0, 0.1, 1.0, 1.1],569                       [0.0, -0.1, 1.0, 0.9], [0.0, 10.0, 1.0, 11.0],570                       [0.0, 10.1, 1.0, 11.1], [0.0, 100.0, 1.0, 101.0]],571                      [[0.0, 0.0, 1.0, 1.0], [0.0, 0.1, 1.0, 1.1],572                       [0.0, -0.1, 1.0, 0.9], [0.0, 10.0, 1.0, 11.0],573                       [0.0, 10.1, 1.0, 11.1], [0.0, 100.0, 1.0,574                                                101.0]]]).astype(np.float32)575    scores = np.array([[[0.9, 0.75, 0.6, 0.95, 0.5, 0.3]],576                       [[0.9, 0.75, 0.6, 0.95, 0.5, 0.3]]]).astype(np.float32)577    max_output_boxes_per_class = np.array([2]).astype(np.int64)578    iou_threshold = np.array([0.5]).astype(np.float32)579    score_threshold = np.array([0.0]).astype(np.float32)580    selected_indices = np.array([[0, 0, 3], [0, 0, 0], [1, 0, 3],581                                 [1, 0, 0]]).astype(np.int64)582    inputs = Input(*[boxes, scores])583    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,584                             iou_threshold))585    self._run(list(zip(inputs, [boxes, scores])))586  def test_non_max_suppression_two_classes(self):587    reset_model(13)588    boxes = np.array([[[0.0, 0.0, 1.0, 1.0], [0.0, 0.1, 1.0, 1.1],589                       [0.0, -0.1, 1.0, 0.9], [0.0, 10.0, 1.0, 11.0],590                       [0.0, 10.1, 1.0, 11.1], [0.0, 100.0, 1.0,591                                                101.0]]]).astype(np.float32)592    scores = np.array([[[0.9, 0.75, 0.6, 0.95, 0.5, 0.3],593                        [0.9, 0.75, 0.6, 0.95, 0.5, 0.3]]]).astype(np.float32)594    max_output_boxes_per_class = np.array([2]).astype(np.int64)595    iou_threshold = np.array([0.5]).astype(np.float32)596    score_threshold = np.array([0.0]).astype(np.float32)597    selected_indices = np.array([[0, 0, 3], [0, 0, 0], [0, 1, 3],598                                 [0, 1, 0]]).astype(np.int64)599    inputs = Input(*[boxes, scores])600    Output(NonMaxSuppression(*inputs, max_output_boxes_per_class,601                             iou_threshold))602    self._run(list(zip(inputs, [boxes, scores])))603  def test_non_zero(self):604    reset_model(13)605    nps = [np.array([[1, 0], [1, 1]], dtype=bool)]606    inputs = Input(*nps)607    Output(NonZero(*inputs))608    self._run(list(zip(inputs, nps)))609  def test_pad(self):610    reset_model(13)611    nps = [612        np.random.randn(1, 2, 3).astype(np.float32),613    ]614    inputs = Input(*nps)615    Output(Pad(*inputs, np.array([0, 0, 1, 0, 0, 2])))616    self._run(list(zip(inputs, nps)))617  def test_pad_5D(self):618    reset_model(13)619    nps = [620        np.random.randn(1, 2, 3, 4, 5).astype(np.float32),621    ]622    inputs = Input(*nps)623    Output(624        Pad(*inputs, np.array([0, 0, 1, 2, 3, 0, 0, 2, 3, 4]),625            np.array([1.0]).astype(np.float32)))626    self._run(list(zip(inputs, nps)))627  def test_pad_reflect(self):628    reset_model(13)629    nps = [630        np.random.randn(1, 2, 3, 4).astype(np.float32),631    ]632    inputs = Input(*nps)633    Output(634        Pad(*inputs,635            np.array([0, 0, 1, 2, 0, 0, 2, 3]),636            np.array([1.0]).astype(np.float32),637            mode="reflect"))638    self._run(list(zip(inputs, nps)))639  def test_reciprocal(self):640    reset_model(13)641    nps = [np.random.randn(1, 10).astype(np.float32)]642    inputs = Input(*nps)643    Output(Reciprocal(inputs[0]))644    self._run(list(zip(inputs, nps)))645  def test_reduce_prod(self):646    reset_model(13)647    nps = [np.random.randn(1, 2, 3).astype(np.float32)]648    inputs = Input(*nps)649    Output(ReduceProd(inputs[0], axes=np.array((1, 2)).astype(np.int64)))650    self._run(list(zip(inputs, nps)))651  def test_reduce_sum(self):652    reset_model(13)653    nps = [np.random.randn(1, 2, 3).astype(np.float32)]654    inputs = Input(*nps)655    Output(ReduceSum(inputs[0], np.array((1, 2)).astype(np.int64)))656    self._run(list(zip(inputs, nps)))657  def test_relu(self):658    reset_model(13)659    nps = [np.random.randn(1, 5).astype(np.float32)]660    inputs = Input(*nps)661    Output(Relu(inputs))662    self._run(list(zip(inputs, nps)))663  def test_elu(self):664    reset_model(13)665    nps = [np.random.randn(1, 5).astype(np.float32)]666    inputs = Input(*nps)667    Output(Elu(inputs))668    self._run(list(zip(inputs, nps)))669  def test_reshape(self):670    reset_model(13)671    nps = [np.random.randn(4,).astype(np.float32)]672    inputs = Input(*nps)673    Output(Reshape(inputs[0], np.array((2, 2)).astype(np.int64)))674    self._run(list(zip(inputs, nps)))675  def test_resize_scales_nearest(self):676    reset_model(13)677    nps = [678        np.array([[[679            [1, 2, 3, 4],680            [5, 6, 7, 8],681        ]]], dtype=np.float32),682        np.array([], dtype=np.float32),683    ]684    inputs = Input(*nps)685    Output(686        Resize(687            *inputs,688            np.array([1.0, 1.0, 0.6, 0.6], dtype=np.float32),689            mode="nearest",690        ))691    self._run(list(zip(inputs, nps)))692  def test_resize_downsample_sizes_linear_pytorch_half_pixel(self):693    reset_model(13)694    nps = [695        np.array([[[696            [1, 2, 3, 4],697            [5, 6, 7, 8],698            [9, 10, 11, 12],699            [13, 14, 15, 16],700        ]]],701                 dtype=np.float32),702        np.array([], dtype=np.float32),703        np.array([], dtype=np.float32),704    ]705    inputs = Input(*nps)706    Output(707        Resize(708            *inputs,709            np.array([1, 1, 3, 1], dtype=np.int64),710            mode='linear',711            # Changed from pytorch_half_pixel(1.8.1) to half_pixel(1.9.0) due to712            # https://github.com/pytorch/pytorch/issues/62237713            coordinate_transformation_mode='half_pixel'))714    self._run(list(zip(inputs, nps)))715  def test_resize_pt_nearest(self):716    reset_model(13)717    nps = [718        np.array([[[[1., 2., 0.], [3., 4., 0.], [0., 0., 0.]]]],719                 dtype=np.float32),720        np.array([], dtype=np.float32),721    ]722    inputs = Input(*nps)723    Output(724        Resize(725            *inputs,726            np.array([1.0, 1.0, 2.0, 2.0], dtype=np.float32),727            mode="nearest",728        ))729    self._run(list(zip(inputs, nps)))730  def test_resize_pt_bilinear(self):731    reset_model(13)732    nps = [733        np.array([[[[1., 2., 0.], [3., 4., 0.], [0., 0., 0.]]]],734                 dtype=np.float32),735        np.array([], dtype=np.float32),736    ]737    inputs = Input(*nps)738    Output(739        Resize(740            *inputs,741            np.array([1.0, 1.0, 2.0, 2.0], dtype=np.float32),742            mode="linear",743        ))744    self._run(list(zip(inputs, nps)))745  def test_resize_pt_bilinear_align_corners(self):746    reset_model(13)747    nps = [748        np.array([[[[1., 2., 0.], [3., 4., 0.], [0., 0., 0.]]]],749                 dtype=np.float32),750        np.array([], dtype=np.float32),751    ]752    inputs = Input(*nps)753    Output(754        Resize(755            *inputs,756            np.array([1.0, 1.0, 2.0, 2.0], dtype=np.float32),757            mode="linear",758            coordinate_transformation_mode="align_corners",759        ))760    self._run(list(zip(inputs, nps)))761  def test_roi_align(self):762    reset_model(13)763    nps = [764        np.array(765            [[[766                [767                    0.2764,768                    0.7150,769                    0.1958,770                    0.3416,771                    0.4638,772                    0.0259,773                    0.2963,774                    0.6518,775                    0.4856,776                    0.7250,777                ],778                [779                    0.9637,780                    0.0895,781                    0.2919,782                    0.6753,783                    0.0234,784                    0.6132,785                    0.8085,786                    0.5324,787                    0.8992,788                    0.4467,789                ],790                [791                    0.3265,792                    0.8479,793                    0.9698,794                    0.2471,795                    0.9336,796                    0.1878,797                    0.4766,798                    0.4308,799                    0.3400,800                    0.2162,801                ],802                [803                    0.0206,804                    0.1720,805                    0.2155,806                    0.4394,807                    0.0653,808                    0.3406,809                    0.7724,810                    0.3921,811                    0.2541,812                    0.5799,813                ],814                [815                    0.4062,816                    0.2194,817                    0.4473,818                    0.4687,819                    0.7109,820                    0.9327,821                    0.9815,822                    0.6320,823                    0.1728,824                    0.6119,825                ],826                [827                    0.3097,828                    0.1283,829                    0.4984,830                    0.5068,831                    0.4279,832                    0.0173,833                    0.4388,834                    0.0430,835                    0.4671,836                    0.7119,837                ],838                [839                    0.1011,840                    0.8477,841                    0.4726,842                    0.1777,843                    0.9923,844                    0.4042,845                    0.1869,846                    0.7795,847                    0.9946,848                    0.9689,849                ],850                [851                    0.1366,852                    0.3671,853                    0.7011,854                    0.6234,855                    0.9867,856                    0.5585,857                    0.6985,858                    0.5609,859                    0.8788,860                    0.9928,861                ],862                [863                    0.5697,864                    0.8511,865                    0.6711,866                    0.9406,867                    0.8751,868                    0.7496,869                    0.1650,870                    0.1049,871                    0.1559,872                    0.2514,873                ],874                [875                    0.7012,876                    0.4056,877                    0.7879,878                    0.3461,879                    0.0415,880                    0.2998,881                    0.5094,882                    0.3727,883                    0.5482,884                    0.0502,885                ],886            ]]],887            dtype=np.float32,888        )889    ]890    inputs = Input(*nps)891    Output(892        RoiAlign(893            inputs[0],894            np.array([[0, 0, 9, 9], [0, 5, 4, 9], [5, 5, 9, 9]],895                     dtype=np.float32),896            np.array([0, 0, 0], dtype=np.int64),897            spatial_scale=1.0,898            output_height=5,899            output_width=5,900            sampling_ratio=2,901        ))902    self._run(list(zip(inputs, nps)))903  def test_scatter_elements_with_axis(self):904    reset_model(13)905    nps = [np.array([[1.0, 2.0, 3.0, 4.0, 5.0]], dtype=np.float32)]906    inputs = Input(*nps)907    Output(908        ScatterElements(inputs[0],909                        np.array([[1, 3]], dtype=np.int64),910                        np.array([[1.1, 2.1]], dtype=np.float32),911                        axis=1))912    self._run(list(zip(inputs, nps)))913  def test_scatter_elements_without_axis(self):914    reset_model(13)915    nps = [np.zeros((3, 3), dtype=np.float32)]916    inputs = Input(*nps)917    Output(918        ScatterElements(919            inputs[0], np.array([[1, 0, 2], [0, 2, 1]], dtype=np.int64),920            np.array([[1.0, 1.1, 1.2], [2.0, 2.1, 2.2]], dtype=np.float32)))921    self._run(list(zip(inputs, nps)))922  def test_scatter_elements_with_negative_axis(self):923    reset_model(13)924    nps = [np.array([[1.0, 2.0, 3.0, 4.0, 5.0]], dtype=np.float32)]925    inputs = Input(*nps)926    Output(927        ScatterElements(inputs[0],928                        np.array([[1, 3]], dtype=np.int64),929                        np.array([[1.1, 2.1]], dtype=np.float32),930                        axis=-1))931    self._run(list(zip(inputs, nps)))932  def test_scatter_elements(self):933    reset_model(13)934    nps = [np.zeros(shape=(10, 7, 7), dtype=np.float32)]935    inputs = Input(*nps)936    Output(937        ScatterElements(inputs[0],938                        np.random.randint(low=0, high=5, size=(5, 7, 7)),939                        np.random.randn(5, 7, 7).astype(np.float32),940                        axis=0))941    self._run(list(zip(inputs, nps)))942  def test_shape(self):943    reset_model(13)944    nps = [np.random.randn(1, 2, 3, 4).astype(np.float32)]945    inputs = Input(*nps)946    Output(Shape(inputs[0]))947    self._run(list(zip(inputs, nps)))948  def test_sigmoid(self):949    reset_model(13)950    nps = [np.random.randn(1, 10).astype(np.float32)]951    inputs = Input(*nps)952    Output(Sigmoid(inputs[0]))953    self._run(list(zip(inputs, nps)))954  def test_slice(self):955    reset_model(13)956    nps = [np.random.randn(5, 5).astype(np.float32)]957    inputs = Input(*nps)958    Output(959        Slice(inputs[0],960              np.array((2, 3)).astype(np.int64),961              np.array((4, 5)).astype(np.int64)))962    self._run(list(zip(inputs, nps)))963  def test_softmax(self):964    reset_model(13)965    nps = [np.random.randn(1, 10).astype(np.float32)]966    inputs = Input(*nps)967    Output(Softmax(inputs[0]))968    self._run(list(zip(inputs, nps)))969  def test_split(self):970    reset_model(13)971    nps = [np.random.randn(1, 10).astype(np.float32)]972    inputs = Input(*nps)973    Output(Split(inputs, split=np.array([2, 8]), axis=1))974    self._run(list(zip(inputs, nps)))975  def test_sqrt(self):976    reset_model(13)977    nps = [np.random.randn(1, 10).astype(np.float32)]978    inputs = Input(*nps)979    Output(Sqrt(inputs[0]))980    self._run(list(zip(inputs, nps)))981  def test_squeeze(self):982    reset_model(13)983    nps = [np.random.randn(1, 10, 1, 1).astype(np.float32)]984    inputs = Input(*nps)985    Output(Squeeze(inputs[0], np.array(([2, 3]))))986    self._run(list(zip(inputs, nps)))987  def test_squeeze_no_axes(self):988    reset_model(13)989    nps = [np.random.randn(1, 10, 1, 1).astype(np.float32)]990    inputs = Input(*nps)991    Output(Squeeze(inputs[0]))992    self._run(list(zip(inputs, nps)))993  def test_topk(self):994    reset_model(13)995    nps = [np.random.randn(1, 2, 3, 4).astype(np.float32)]996    inputs = Input(*nps)997    Output(TopK(inputs[0], np.asarray([3])))998    self._run(list(zip(inputs, nps)))999  def test_topk_attrs(self):1000    reset_model(13)1001    nps = [np.random.randn(1, 1, 5, 1).astype(np.float32)]1002    inputs = Input(*nps)1003    Output(TopK(inputs[0], np.asarray([3]), axis=2, largest=0, sorted=1))1004    self._run(list(zip(inputs, nps)))1005  def test_transpose(self):1006    reset_model(13)1007    nps = [np.random.randn(1, 2, 3, 4).astype(np.float32)]1008    inputs = Input(*nps)1009    Output(Transpose(inputs[0], perm=[0, 2, 3, 1]))1010    self._run(list(zip(inputs, nps)))1011  def test_transpose_no_perm(self):1012    reset_model(13)1013    nps = [np.random.randn(1, 2, 3, 4).astype(np.float32)]1014    inputs = Input(*nps)1015    Output(Transpose(inputs[0]))1016    self._run(list(zip(inputs, nps)))1017  def test_unsqueeze(self):1018    reset_model(13)1019    nps = [np.random.randn(1, 2).astype(np.float32)]1020    inputs = Input(*nps)1021    Output(Unsqueeze(inputs[0], np.array(([2, 3]))))1022    self._run(list(zip(inputs, nps)))1023if __name__ == '__main__':...

model.py

Source:model.py

1import numpy as np2def map_values(data, mapping):3    # maps values of numpy array using a mapping dictionary4    return np.vectorize(mapping.get)(data)5def sign(x):6    # implements np.sign function as sets random choice of 1 or -1 for 0 results7    x_out = np.sign(x)8    if x_out == 0:9        x_out = np.random.choice([-1, 1])10    return x_out11# load data into training and test numpy array, and dictionary into dict()12directory = 'data/'13files = [directory + file for file in ['pa3dictionary.txt', 'pa3test.txt', 'pa3train.txt']]14row_to_word = dict(enumerate(list(np.loadtxt(files[0], dtype='str'))))15train = np.loadtxt(files[1], dtype='int', delimiter=' ')16test = np.loadtxt(files[2], dtype='int', delimiter=' ')17num_features = train.shape[1] - 118# map 2 / 1 labels to -1 / 119mapping = {2: -1, 1: 1}20train[:, -1] = map_values(train[:, -1], mapping)21test[:, -1] = map_values(test[:, -1], mapping)22# simple perceptron class23class SimplePerceptron:24    def __init__(self):25        self.num_features = None26        self.w = None27    def fit(self, labeled_data, reset_model=True):28        # fit model to labeled_data, where29        #   labeled_data is 2d array, with columns as features and rows as samples30        #   final column is label, must be -1 or 131        X = labeled_data[:, :-1]32        y = labeled_data[:, -1]33        if reset_model:34            self.num_features = X.size[1]35            self.w = np.zeros(self.num_features + 1)36        for i in range(y.size):37            x = np.append(X[i], [1])  # append 1 to the x, for bias term38            yhat = self.predict(x)  # estimate class using current weights39            if y[i] != yhat:  # if incorrect decision40                self.w = np.add(self.w, np.multiply(y[i], x))  # update weights41    def predict(self, x):42        return sign(np.dot(x, self.w))43    def test(self, labeled_data):44        # test classification of model using labeled_data, where45        #   labeled_data is 2d array, with columns as features and rows as samples46        #   final column is label, must be -1 or 147        X = labeled_data[:, :-1]48        y = labeled_data[:, -1]49        num_errors = 050        for i in range(y.size):51            x = np.append(X[i], [1])  # append 1 to the x, for bias term52            yhat = self.predict(x)  # estimate class53            if y[i] != yhat:  # if incorrect decision54                num_errors += 155        return num_errors / y.size56# train model over 5 passes of the training data and report training & test error each time57simple_perceptron = SimplePerceptron()58print('## Simple Perceptron ##')59num_passes = 560for i in range(num_passes):61    print('# Pass ' + str(i+1) + ' #')62    simple_perceptron.fit(train, reset_model=(i == 0))  # fit model and reset_model when i==063    print('Training error: ' + '{:.4f}'.format(simple_perceptron.test(train)))64    print('Testing error:  ' + '{:.4f}'.format(simple_perceptron.test(test)))65    print('')66print('')67print('')68# kernelised perceptron class69class KernelisedPerceptron:70    def __init__(self):71        self.num_samples = None72        self.num_features = None73        self.alpha = None74        self.X_trained = None75        self.kernel = None76    def fit(self, labeled_data, reset_model=True, kernel=np.dot):77        # fit model to labeled_data using kernel, where78        #   labeled_data is 2d array, with columns as features and rows as samples79        #   final column is label, must be -1 or 180        X = labeled_data[:, :-1]81        y = labeled_data[:, -1]82        if reset_model:83            self.kernel = kernel84            self.X_trained = X85            self.num_samples, self.num_features = X.shape86            self.alpha = np.zeros(self.num_samples)87        for j in range(self.num_samples):88            yhat = self.predict(X[j, :])89            if y[j] != yhat:90                self.alpha[j] += y[j]91    def predict(self, x):92        # classify sample x using trained model93        yhat = 094        for i in range(self.num_samples):95            yhat += self.alpha[i] * self.kernel(self.X_trained[i, :], x)96        return sign(yhat)97    def test(self, labeled_data):98        # test classification using model on labeled_data, where99        #   labeled_data is 2d array, with columns as features and rows as samples100        #   final column is label, must be -1 or 1101        #   returns error rate102        num_errors = 0103        X = labeled_data[:, :-1]104        y = labeled_data[:, -1]105        for j in range(self.num_samples):106            yhat = self.predict(X[j, :])107            if y[j] != yhat:108                num_errors += 1109        return num_errors / self.num_samples110# kernel functions as defined in exercise111def kernel_exp(x1, x2):112    return np.exp(-np.linalg.norm(x1-x2) / 20)113def kernel_poly(x1, x2):114    return (np.dot(x1, x2) + 10) ** 2115kernel_perceptron = KernelisedPerceptron()116kernels = [np.dot, kernel_exp, kernel_poly]117messages = ['## Kernalised - dot product ##',118            '## Kernalised - exponential ##',119            '## Kernalised - polynomial ##']120# for each kernel, train model over 5 passes of the training data and report training & test error each time121num_passes = 5122for message, kernel in zip(messages, kernels):123    print(message)124    for i in range(num_passes):125        print('# Pass ' + str(i+1) + ' #')126        kernel_perceptron.fit(train, reset_model=(i == 0), kernel=kernel)  # fit model and reset_model when i==0127        print('Training error: ' + '{:.5f}'.format(kernel_perceptron.test(train)))128        print('Testing error:  ' + '{:.5f}'.format(kernel_perceptron.test(test)))129        print('')130    print('')...

Find_Best_Learning_Rate.py

Source:Find_Best_Learning_Rate.py

1import os2from tensorflow.python.keras.optimizers import Adam3from tensorflow.python.keras.backend import get_session4def reset_model(model):5    print('Reset model')6    session = get_session()7    for i, layer in enumerate(model.layers):8        if hasattr(layer, 'kernel_initializer'):9            model.layers[i].kernel.initializer.run(session=session)10    return model11class Find_Best_Learning_Rate(object):12    def __init__(self, train_generator=None, validation_generator=None, Model_val=None, epochs=2, learning_rate=0, upper_bound=1, scale=1.1,reset_model=False,13                 out_path=os.path.join('.','Learning_rates'),metrics=['accuracy'], optimizer=Adam, loss='categorical_crossentropy',num_workers=10, steps_per_epoch=None):14        self.steps_per_epoch = self.steps_per_epoch15        if steps_per_epoch is None:16            self.steps_per_epoch = len(train_generator)17        self.num_workers = num_workers18        self.reset_model = reset_model19        self.loss = loss20        self.train_generator = train_generator21        self.validation_generator = validation_generator22        self.epochs = epochs23        self.Model_val = Model_val24        self.out_path = out_path25        self.metrics = metrics26        self.optimizer = optimizer27        if not os.path.exists(out_path):28            os.makedirs(out_path)29        self.run_over_learning_rates(learning_rate=learning_rate, upper_bound=upper_bound, scale=scale)30    def run_for_learning_rate(self, learning_rate):31        out_path = os.path.join(self.out_path, str(learning_rate) + '.txt')32        if os.path.exists(out_path):33            return None # Do not redo a run34        if self.reset_model:35            self.Model_val = reset_model(self.Model_val)36        optimizer = self.optimizer(lr=learning_rate)37        self.Model_val.compile(optimizer, loss=self.loss, metrics=self.metrics)38        history = self.Model_val.fit_generator(generator=self.train_generator, workers=self.num_workers, use_multiprocessing=False,39                                               max_queue_size=200,shuffle=True, epochs=self.epochs, initial_epoch=0,40                                               validation_data=self.validation_generator, steps_per_epoch=self.steps_per_epoch)41        fid = open(out_path,'w+')42        for key in history.history.keys():43            if key.find('val') == 0: # Only get the validation data44                fid.write(key + ',')45                fid.write(str(history.history[key])+'\n')46        fid.close()47        return None48    def run_over_learning_rates(self, learning_rate, upper_bound, scale):49        while learning_rate < upper_bound:...

test_sentiment_classifier.py

Source:test_sentiment_classifier.py

...15      pos['training']16    ])17    c = 2 ** 718    classifier.c = c19    classifier.reset_model()20    n_er = self.validate(classifier, neg['validation'], 'negative')21    p_er = self.validate(classifier, pos['validation'], 'positive')22    total = Fraction(n_er.numerator + p_er.numerator,23                     n_er.denominator + p_er.denominator)24    print(total)25    self.assertLess(total, 0.35)26  def test_validate_itself(self):27    """yields a zero error when it uses itself"""28    classifier = SentimentClassifier.build([29      'data/rt-polaritydata/rt-polarity.neg',30      'data/rt-polaritydata/rt-polarity.pos'31    ])32    c = 2 ** 733    classifier.c = c34    classifier.reset_model()35    n_er = self.validate(classifier,36                         'data/rt-polaritydata/rt-polarity.neg',37                         'negative')38    p_er = self.validate(classifier,39                         'data/rt-polaritydata/rt-polarity.pos',40                         'positive')41    total = Fraction(n_er.numerator + p_er.numerator,42                     n_er.denominator + p_er.denominator)43    print(total)44    self.assertEqual(total, 0)45  def validate(self, classifier, file, sentiment):46    total = 047    misses = 048    with(io.open(file, errors='ignore')) as f:...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.