| /external/google-fruit/extras/benchmark/tables/ |
| D | fruit_wiki.yml | 10 dimension: "num_classes"
81 num_classes: 100
107 num_classes: 250
133 num_classes: 1000
159 num_classes: 100
185 num_classes: 250
211 num_classes: 1000
237 num_classes: 100
263 num_classes: 250
289 num_classes: 1000
[all …]
|
| /external/google-fruit/extras/benchmark/suites/ |
| D | fruit_full.yml | 24 num_classes: &num_classes
51 num_classes: *num_classes
62 num_classes: *num_classes
79 num_classes: *num_classes
91 num_classes: *num_classes
|
| D | debug.yml | 64 num_classes:
78 num_classes:
96 num_classes:
110 num_classes:
130 num_classes:
142 num_classes:
159 num_classes:
171 num_classes:
|
| D | fruit_mostly_full.yml | 24 num_classes: &num_classes
48 num_classes: *num_classes
59 num_classes: *num_classes
|
| D | simple_di_full.yml | 27 num_classes: &num_classes
54 num_classes: *num_classes
67 num_classes: *num_classes
|
| D | fruit_quick.yml | 24 num_classes: &num_classes
49 num_classes: *num_classes
|
| D | simple_di_mostly_full.yml | 27 num_classes: &num_classes
52 num_classes:
66 num_classes:
|
| /external/tensorflow/tensorflow/python/ops/ |
| D | confusion_matrix.py | 94 num_classes=None, argument
106 If `num_classes` is `None`, then `num_classes` will be set to one plus the
108 start at 0. For example, if `num_classes` is 3, then the possible labels
131 num_classes: The possible number of labels the classification task can
149 (predictions, labels, num_classes, weights)) as name:
167 if num_classes is None:
168 num_classes = math_ops.maximum(math_ops.reduce_max(predictions),
171 num_classes_int64 = math_ops.cast(num_classes, dtypes.int64)
187 shape = array_ops.stack([num_classes, num_classes])
202 num_classes=None, argument
[all …]
|
| D | nn_test.py | 527 def _GenerateTestData(self, num_classes, dim, batch_size, num_true, labels, argument
534 num_classes: An int. The number of embedding classes in the test case.
539 sampled: A list of indices in [0, num_classes).
545 of shape [num_classes, dim]
547 of shape [num_classes].
557 weights = np.random.randn(num_classes, dim).astype(np.float32)
558 biases = np.random.randn(num_classes).astype(np.float32)
613 num_classes = 5
618 low=0, high=num_classes, size=batch_size * num_true)
621 num_classes=num_classes,
[all …]
|
| /external/tensorflow/tensorflow/core/kernels/ |
| D | in_topk_op_gpu.cu.cc | 40 const T* __restrict__ predictions, // dims: [ num_targets x num_classes ] in ComputePredictionMaskKernel()
42 int64* __restrict__ mask, // dims: [ num_targets x num_classes ] in ComputePredictionMaskKernel()
43 int num_targets, int num_classes) { in ComputePredictionMaskKernel() argument
44 GPU_1D_KERNEL_LOOP(i, num_targets * num_classes) { in ComputePredictionMaskKernel()
45 const int batch_index = i / num_classes; in ComputePredictionMaskKernel()
48 if (!FastBoundsCheck(target_idx, num_classes)) { in ComputePredictionMaskKernel()
55 ldg(predictions + batch_index * num_classes + target_idx); in ComputePredictionMaskKernel()
96 const Eigen::Index num_classes = predictions.dimension(1); in operator ()() local
99 context, num_targets * num_classes < std::numeric_limits<int>::max(), in operator ()()
103 if (num_targets == 0 || num_classes == 0) { in operator ()()
[all …]
|
| D | multinomial_op.cc | 51 int num_classes, int num_samples,
77 int num_classes, int num_samples, in operator ()()
86 auto DoWork = [ctx, num_samples, num_classes, &gen, &output, &logits]( in operator ()()
99 ctx->allocate_temp(DT_DOUBLE, TensorShape({num_classes}), in operator ()()
107 for (int64_t j = 0; j < num_classes; ++j) { in operator ()()
119 for (int64_t j = 0; j < num_classes; ++j) { in operator ()()
127 const double* cdf_end = cdf.data() + num_classes; in operator ()()
137 50 * (num_samples * std::log(num_classes) / std::log(2) + num_classes); in operator ()()
176 const int num_classes = static_cast<int>(logits_t.dim_size(1)); in DoCompute() local
177 OP_REQUIRES(ctx, num_classes > 0, in DoCompute()
[all …]
|
| D | ctc_decoder_ops.cc | 212 errors::InvalidArgument("num_classes cannot exceed max int")); in Compute()
213 const int num_classes = static_cast<const int>(num_classes_raw); in Compute() local
219 input_list_t.emplace_back(inputs_t.data() + t * batch_size * num_classes, in Compute()
220 batch_size, num_classes); in Compute()
228 (blank_index_ < 0) ? num_classes + blank_index_ : blank_index_; in Compute()
229 OP_REQUIRES(ctx, FastBoundsCheck(blank_index, num_classes), in Compute()
231 -num_classes, " and ", num_classes - 1, in Compute()
256 const int64_t kCostPerUnit = 50 * max_time * num_classes; in Compute()
320 errors::InvalidArgument("num_classes cannot exceed max int")); in Compute()
321 const int num_classes = static_cast<const int>(num_classes_raw); in Compute() local
[all …]
|
| D | multinomial_op_gpu.cu.cc | 42 __global__ void MultinomialKernel(int32 nthreads, const int32 num_classes, in MultinomialKernel() argument
48 const int maxima_idx = index / num_classes; in MultinomialKernel()
52 static_cast<UnsignedOutputType>(index % num_classes)); in MultinomialKernel()
64 int num_classes, int num_samples, in operator ()()
76 bsc.set(2, num_classes); in operator ()()
80 boc.set(2, num_classes); in operator ()()
104 /*in_dim1=*/num_classes, /*in_dim2=*/1, /*out_rank=*/1, in operator ()()
110 const int32 work_items = batch_size * num_samples * num_classes; in operator ()()
115 num_classes, num_samples, scores.data(), maxima.data(), output.data())); in operator ()()
|
| D | ctc_loss_op.cc | 121 errors::InvalidArgument("num_classes cannot exceed max int")); in Compute()
122 const int num_classes = static_cast<const int>(num_classes_raw); in Compute() local
198 input_list_t.emplace_back(inputs_t.data() + t * batch_size * num_classes, in Compute()
199 batch_size, num_classes); in Compute()
201 gradient_t.data() + t * batch_size * num_classes, batch_size, in Compute()
202 num_classes); in Compute()
207 // Assumption: the blank index is num_classes - 1 in Compute()
208 ctc::CTCLossCalculator<T> ctc_loss_calculator(num_classes - 1, 0); in Compute()
295 errors::InvalidArgument("num_classes cannot exceed max int")); in Compute()
298 const int num_classes = static_cast<const int>(num_classes_raw); in Compute() local
[all …]
|
| /external/tensorflow/tensorflow/python/kernel_tests/random/ |
| D | multinomial_op_test.py | 150 logits: Numpy ndarray of shape [batch_size, num_classes].
152 sampler: A sampler function that takes (1) a [batch_size, num_classes]
156 Frequencies from sampled classes; shape [batch_size, num_classes].
163 batch_size, num_classes = logits.shape
169 self.assertLess(max(cnts.keys()), num_classes)
173 for k in range(num_classes)]
199 with self.assertRaisesOpError("num_classes should be positive"):
212 def native_op_vs_composed_ops(batch_size, num_classes, num_samples, num_iters): argument
214 shape = [batch_size, num_classes]
243 for num_classes in [10000, 100000]:
[all …]
|
| /external/tensorflow/tensorflow/python/keras/utils/ |
| D | np_utils.py | 22 def to_categorical(y, num_classes=None, dtype='float32'): argument
29 (integers from 0 to num_classes).
30 num_classes: total number of classes. If `None`, this would be inferred
40 >>> a = tf.keras.utils.to_categorical([0, 1, 2, 3], num_classes=4)
71 if not num_classes:
72 num_classes = np.max(y) + 1
74 categorical = np.zeros((n, num_classes), dtype=dtype)
76 output_shape = input_shape + (num_classes,)
|
| /external/tensorflow/tensorflow/core/util/ctc/ |
| D | ctc_loss_calculator.h | 99 int num_classes, const Vector& seq_len,
131 auto num_classes = inputs[0].cols(); in CalculateLoss() local
144 if (inputs[t].cols() != num_classes) { in CalculateLoss()
146 " to be: ", num_classes, in CalculateLoss()
169 batch_size, num_classes, seq_len, labels, &max_u_prime, &l_primes); in CalculateLoss()
175 auto ComputeLossAndGradients = [this, num_classes, &labels, &l_primes, in CalculateLoss()
204 Matrix y(num_classes, seq_len(b)); in CalculateLoss()
214 // y, prob are in num_classes x seq_len(b) in CalculateLoss()
270 max_seq_len * num_classes * in CalculateLoss()
272 max_seq_len * 2 * (2 * num_classes + 1) * in CalculateLoss()
[all …]
|
| D | ctc_beam_search_test.cc | 109 const int num_classes = 6; in ctc_beam_search_decoding_with_and_without_dictionary() local
114 tensorflow::ctc::CTCBeamSearchDecoder<T> decoder(num_classes, 10 * top_paths, in ctc_beam_search_decoding_with_and_without_dictionary()
120 dictionary_decoder(num_classes, top_paths, &dictionary_scorer); in ctc_beam_search_decoding_with_and_without_dictionary()
124 T input_data_mat[timesteps][batch_size][num_classes] = { in ctc_beam_search_decoding_with_and_without_dictionary()
134 for (int c = 0; c < num_classes; ++c) { in ctc_beam_search_decoding_with_and_without_dictionary()
163 inputs.emplace_back(&input_data_mat[t][0][0], batch_size, num_classes); in ctc_beam_search_decoding_with_and_without_dictionary()
199 const int num_classes = 6; in ctc_beam_search_decoding_all_beam_elements_have_finite_scores() local
204 tensorflow::ctc::CTCBeamSearchDecoder<T> decoder(num_classes, top_paths, in ctc_beam_search_decoding_all_beam_elements_have_finite_scores()
209 T input_data_mat[timesteps][batch_size][num_classes] = { in ctc_beam_search_decoding_all_beam_elements_have_finite_scores()
221 inputs.emplace_back(&input_data_mat[t][0][0], batch_size, num_classes); in ctc_beam_search_decoding_all_beam_elements_have_finite_scores()
[all …]
|
| D | ctc_decoder.h | 46 CTCDecoder(int num_classes, int batch_size, bool merge_repeated) in CTCDecoder() argument
47 : num_classes_(num_classes), in CTCDecoder()
48 blank_index_(num_classes - 1), in CTCDecoder()
64 int num_classes() { return num_classes_; } in num_classes() function
79 CTCGreedyDecoder(int num_classes, int batch_size, bool merge_repeated) in CTCGreedyDecoder() argument
80 : CTCDecoder<T>(num_classes, batch_size, merge_repeated) {} in CTCGreedyDecoder()
|
| /external/tensorflow/tensorflow/python/keras/ |
| D | testing_utils.py | 60 num_classes, argument
68 num_classes: Integer, number of classes for the data and targets.
77 templates = 2 * num_classes * np.random.random((num_classes,) + input_shape)
78 y = np.random.randint(0, num_classes, size=(num_sample,))
422 def get_small_sequential_mlp(num_hidden, num_classes, input_dim=None): argument
428 activation = 'sigmoid' if num_classes == 1 else 'softmax'
429 model.add(layers.Dense(num_classes, activation=activation))
433 def get_small_functional_mlp(num_hidden, num_classes, input_dim): argument
436 activation = 'sigmoid' if num_classes == 1 else 'softmax'
437 outputs = layers.Dense(num_classes, activation=activation)(outputs)
[all …]
|
| /external/tensorflow/tensorflow/python/ops/numpy_ops/integration_test/benchmarks/ |
| D | tf_numpy_mlp.py | 20 NUM_CLASSES = 3 variable
32 def __init__(self, num_classes=NUM_CLASSES, input_size=INPUT_SIZE, argument
36 self.w2 = np.random.uniform(size=[hidden_units, num_classes]).astype(
40 self.b2 = np.random.uniform(size=[1, num_classes]).astype(
|
| D | numpy_mlp.py | 18 NUM_CLASSES = 3 variable
30 def __init__(self, num_classes=NUM_CLASSES, input_size=INPUT_SIZE, argument
34 self.w2 = np.random.uniform(size=[hidden_units, num_classes]).astype(
38 self.b2 = np.random.uniform(size=[1, num_classes]).astype(
|
| /external/ComputeLibrary/src/core/CPP/kernels/ |
| D | CPPBoxWithNonMaximaSuppressionLimitKernel.cpp | 201 const int num_classes = _scores_in->info()->dimension(0); in run_nmslimit() local
204 std::vector<std::vector<int>> keeps(num_classes); in run_nmslimit()
207 std::vector<std::vector<T>> in_scores(num_classes, std::vector<T>(scores_count)); in run_nmslimit()
210 for(int j = 0; j < num_classes; ++j) in run_nmslimit()
220 const int j_start = (num_classes == 1 ? 0 : 1); in run_nmslimit()
221 for(int j = j_start; j < num_classes; ++j) in run_nmslimit()
276 for(int j = 1; j < num_classes; ++j) in run_nmslimit()
296 for(int j = j_start; j < num_classes; ++j) in run_nmslimit()
323 for(int j = 0; j < num_classes; ++j) in run_nmslimit()
329 …rpret_cast<uint32_t *>(_keeps_size->ptr_to_element(Coordinates(j + b * num_classes))) = keeps[j].s… in run_nmslimit()
[all …]
|
| /external/tensorflow/tensorflow/lite/kernels/ctc/ |
| D | ctc_decoder.h | 43 CTCDecoder(int num_classes, int batch_size, bool merge_repeated) in CTCDecoder() argument
44 : num_classes_(num_classes), in CTCDecoder()
45 blank_index_(num_classes - 1), in CTCDecoder()
61 int num_classes() { return num_classes_; } in num_classes() function
74 CTCGreedyDecoder(int num_classes, int batch_size, bool merge_repeated) in CTCGreedyDecoder() argument
75 : CTCDecoder(num_classes, batch_size, merge_repeated) {} in CTCGreedyDecoder()
|
| /external/tensorflow/tensorflow/tools/android/test/src/org/tensorflow/demo/ |
| D | TensorFlowYoloDetector.java | 37 private static final int NUM_CLASSES = 20; field in TensorFlowYoloDetector
171 new float[gridWidth * gridHeight * (NUM_CLASSES + 5) * NUM_BOXES_PER_BLOCK]; in recognizeImage()
191 (gridWidth * (NUM_BOXES_PER_BLOCK * (NUM_CLASSES + 5))) * y in recognizeImage()
192 + (NUM_BOXES_PER_BLOCK * (NUM_CLASSES + 5)) * x in recognizeImage()
193 + (NUM_CLASSES + 5) * b; in recognizeImage()
212 final float[] classes = new float[NUM_CLASSES]; in recognizeImage()
213 for (int c = 0; c < NUM_CLASSES; ++c) { in recognizeImage()
218 for (int c = 0; c < NUM_CLASSES; ++c) { in recognizeImage()
|