Randomness araka ny famaritana ny fiteny malaza indrindra amin'ny Machine Learning

Fanamarihana ny votoatin'ny Ny fametrahana Ny tf.random.Generator kilasy Ny famoronana fehezan-dalàna tsy miankina Ny fifandraisana amin'ny tf.function Ny fifandraisana amin'ny sehatry ny fizarana Ny famonjena ny generator Tsy misy ny RNGs Ny algorithm Ny jeneraly Ny fitaovana XLA Ny TensorFlow dia manome vondrona mpamorona isan-karazany (RNG), ao amin'ny Ity antontan-taratasy ity dia mamaritra ny fomba ahafahanao mifehy ny solosaina famokarana isan-karazany, ary ny fomba mifandray amin'ireo solosaina ireo amin'ny rafitra hafa. tf.random Fanamarihana: Tsy azo antoka fa mifanaraka amin'ny dikan'ny TensorFlow ireo isan-karazany. jereo: Fifanarahana amin'ny dikan Fanamarihana: Tsy azo antoka fa mifanaraka amin'ny dikan'ny TensorFlow ireo isan-karazany. jereo: Fifanarahana amin'ny dikan Fifanarahana amin'ny version TensorFlow dia manome fomba roa ho an'ny fanaraha-maso ny dingana famokarana isan-karazany: Amin'ny alalan'ny fampiasana mazava ny tf.random.Generator objekts. Ny tsirairay toy izany dia mitazona ny toe-javatra (ao amin'ny tf.Variable) izay hiova aorian'ny famoronana isan-karazany. Amin'ny alàlan'ny fonosam-pifandraisana tsy misy fananana, toy ny tf.random.stateless_uniform. Ny fangatahana ireo endri-javatra ireo amin'ny arguments mitovy (anisan'izany ny toetra) ary amin'ny fitaovana mitovy dia hiteraka vokatra mitovy foana. Fanamarihana: Ny RNG taloha avy amin'ny TF 1.x toy ny tf.random.uniform sy tf.random.normal dia mbola tsy voavonjy, fa voavonjy mafy. Ny RNGs taloha avy amin'ny TF 1.x toy ny Ary Mbola tsy voafehy izy ireo, fa voafehy mafy. Warning: tf.random.uniform tf.random.normal Ny fametrahana import tensorflow as tf # Creates some virtual devices (cpu:0, cpu:1, etc.) for using distribution strategy physical_devices = tf.config.list_physical_devices("CPU") tf.config.experimental.set_virtual_device_configuration( physical_devices[0], [ tf.config.experimental.VirtualDeviceConfiguration(), tf.config.experimental.VirtualDeviceConfiguration(), tf.config.experimental.VirtualDeviceConfiguration() ]) 2024-08-15 01:43:41.157432: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered 2024-08-15 01:43:41.178819: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered 2024-08-15 01:43:41.185039: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR I0000 00:00:1723686223.758551 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.762466 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.765643 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.769228 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.780976 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.784469 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.787369 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.790779 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.794250 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.797851 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.800627 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686223.804177 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 Ny Ny kilasy Ny mpamatsy ny Generator Ny mpamatsy ny Generator Ny Ity kilasy ity dia ampiasaina amin'ny toe-javatra izay tianao ny hiteraka vokatra samihafa amin'ny antso RNG tsirairay. Izany dia azo ampiasaina isaky ny fametrahana isan-karazany. satria ny toe-javatra dia Mankafy ny fitaovana rehetra omena amin'ny toy ny fanamafisana tsotra, fanaraha-maso automatique sy ny fiarovana ny fitaovana. tf.random.Generator tf.Variable tf.Variable tf.Variable Azonao atao ny mahazo amin'ny alalan'ny famoronana fitaovana avy amin'ny kilasy na antso Mba hahazoana ny Global Generator default: tf.random.Generator tf.random.get_global_generator() g1 = tf.random.Generator.from_seed(1) print(g1.normal(shape=[2, 3])) g2 = tf.random.get_global_generator() print(g2.normal(shape=[2, 3])) I0000 00:00:1723686225.040869 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.043016 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.045041 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.047115 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.049184 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.051188 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.053162 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.055160 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.057105 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.059113 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.061029 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 tf.Tensor( [[ 0.43842277 -0.53439844 -0.07710262] [ 1.5658045 -0.1012345 -0.2744976 ]], shape=(2, 3), dtype=float32) tf.Tensor( [[ 1.3061213 0.6299361 0.52625704] [ 1.3733886 0.29277426 -0.7945693 ]], shape=(2, 3), dtype=float32) I0000 00:00:1723686225.063243 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.101904 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.103956 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.105919 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.107957 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.109922 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.111909 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.113815 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.115823 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.117783 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.120279 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.122678 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1723686225.125117 55391 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 Misy fomba maro ahafahana mamorona lohahevitra. Ny tsotra indrindra dia , araka ny voalaza etsy ambony, izay mamorona generator avy amin'ny taranaka. Ny taranaka dia isan-karazany tsy manan-karena. Ankoatra izany, misy ihany koa ny antony tsy ara-dalàna izay ny algorithm RNG izay hampiasa ity generator ity: Generator.from_seed from_seed alg g1 = tf.random.Generator.from_seed(1, alg='philox') print(g1.normal(shape=[2, 3])) tf.Tensor( [[ 0.43842277 -0.53439844 -0.07710262] [ 1.5658045 -0.1012345 -0.2744976 ]], shape=(2, 3), dtype=float32) Jereo ny Ny ampahany etsy ambany dia ahitana fampahalalana bebe kokoa momba izany. Ny algorithm Ny fomba iray hafa ahafahana mamorona generator dia amin'ny Ny generator natao amin'ity fomba ity dia hanomboka amin'ny toe-javatra tsy determinist, miankina amin'ny ohatra, ny fotoana sy ny OS. Generator.from_non_deterministic_state g = tf.random.Generator.from_non_deterministic_state() print(g.normal(shape=[2, 3])) tf.Tensor( [[ 0.9104948 -0.23143363 -0.09841432] [-0.91448975 0.1579936 1.3923475 ]], shape=(2, 3), dtype=float32) Misy fomba hafa ihany koa mba hamoronana generators, toy ny avy amin'ny fanjakana mazava, izay tsy voatahiry amin'ity torohevitra ity. Rehefa mampiasa mba hahazoana ny Global Generator, mila mitandrina momba ny fitaovana toerana. Ny Global Generator dia naorina (tsy avy amin'ny tsy-deterministic fanjakana) amin'ny fotoana voalohany dia antsoina ary napetraka ao amin'ny fitaovana default amin'izany antsoina izany. Noho izany, ohatra, raha ny toerana voalohany Ao anatin'ny a Ao anatin'ny sehatra, ny Global Generator dia halefa ao amin'ny GPU, ary amin'ny fampiasana ny Global Generator avy amin'ny CPU dia hisy kopia GPU-to-CPU. tf.random.get_global_generator tf.random.get_global_generator tf.random.get_global_generator tf.device("gpu") Misy ihany koa ny andraikitra Na izany aza, tokony hampiasaina amin'ny fahasalamana ity endri-javatra ity, satria ny generator global taloha dia mety ho voasambotra amin'ny (ho toy ny fanehoan-kevitra malemy), ary ny fanovana azy dia hitondra azy ho fako voaangona, manimba ny Ny fomba tsara kokoa hamerina ny Global Generator dia mampiasa ny iray amin'ireo "reset" asa toy ny , izay tsy hamorona zavatra vaovao amin'ny generator. tf.random.set_global_generator tf.function tf.function Generator.reset_from_seed g = tf.random.Generator.from_seed(1) print(g.normal([])) print(g.normal([])) g.reset_from_seed(1) print(g.normal([])) tf.Tensor(0.43842277, shape=(), dtype=float32) tf.Tensor(1.6272374, shape=(), dtype=float32) tf.Tensor(0.43842277, shape=(), dtype=float32) Ny famoronana fehezan-dalàna tsy miankina Ao amin'ny fampiharana maro dia ilaina ny fantsona isan-karazany tsy miankina amin'ny isan-karazany, tsy miankina amin'ny fahatsapana fa tsy hifandray izy ireo ary tsy manana fifandraisana azo jerena amin'ny antontan-taratasy. mba hamoronana generators maromaro izay azo antoka fa tsy miankina amin'ny iray amin'ny iray hafa (ohatra mamokatra herinaratra tsy miankina). Generator.split g = tf.random.Generator.from_seed(1) print(g.normal([])) new_gs = g.split(3) for new_g in new_gs: print(new_g.normal([])) print(g.normal([])) tf.Tensor(0.43842277, shape=(), dtype=float32) tf.Tensor(2.536413, shape=(), dtype=float32) tf.Tensor(0.33186463, shape=(), dtype=float32) tf.Tensor(-0.07144657, shape=(), dtype=float32) tf.Tensor(-0.79253083, shape=(), dtype=float32) Izany dia hanova ny toe-javatra amin'ny generator izay antsoina hoe ( amin'ny ohatra etsy ambony), mitovy amin'ny fomba RNG toy ny Ankoatra ny tsy miankina amin'ny hafa, ny solosaina vaovao ( ) dia azo antoka ihany koa fa tsy miankina amin'ny taloha ( Ny split g normal new_gs g Ny fametrahana solosaina vaovao dia mahasoa ihany koa raha tianao ny hahazo antoka fa ny solosaina ampiasainao dia ao amin'ny fitaovana mitovy amin'ny solosaina hafa, mba hisorohana ny fiantraikany amin'ny famindrana fitaovana. ohatra: with tf.device("cpu"): # change "cpu" to the device you want g = tf.random.get_global_generator().split(1)[0] print(g.normal([])) # use of g won't cause cross-device copy, unlike the global generator tf.Tensor(-0.66787744, shape=(), dtype=float32) Fanamarihana: Amin'ny teoria, azonao atao ny mampiasa mpanamboatra toy ny from_seed fa tsy ny fisarahana eto mba hahazoana mpamokatra vaovao, fa amin'izany dia ho very ny antoka fa ny mpamokatra vaovao dia tsy miankina amin'ny mpamokatra maneran-tany. Fanamarihana: Amin'ny teoria, azonao atao ny mampiasa mpanamboatra toy ny from_seed fa tsy ny fisarahana eto mba hahazoana mpamokatra vaovao, fa amin'izany dia ho very ny antoka fa ny mpamokatra vaovao dia tsy miankina amin'ny mpamokatra maneran-tany. Azonao atao ny manao ny fizarana recursively, miantso Tsy misy fetra (tsy misy fetra) amin'ny habetsahan'ny recursions. split Ny fifandraisana amin'ny tf.function Manaraka ny fitsipika mitovy amin'ny Rehefa ampiasaina amin'ny Izany dia ahitana endrika telo. tf.random.Generator tf.Variable tf.function Ny famoronana generators ivelany Ny fandefasana Ny fandefasana Azonao atao ny mampiasa generator izay naorina ivelany. tf.function g = tf.random.Generator.from_seed(1) @tf.function def foo(): return g.normal([]) print(foo()) tf.Tensor(0.43842277, shape=(), dtype=float32) Ny mpampiasa dia mila mahazo antoka fa mbola velona ny lohahevitra generator (tsy voaangona) rehefa antsoina ny asa. Ny famoronana generators ao anatiny Ny fandefasana Ny fandefasana Ny famoronana generators ao anatin'ny a Mety hitranga amin'ny fotoana voalohany amin'ny asa. tf.function g = None @tf.function def foo(): global g if g is None: g = tf.random.Generator.from_seed(1) return g.normal([]) print(foo()) print(foo()) tf.Tensor(0.43842277, shape=(), dtype=float32) tf.Tensor(1.6272374, shape=(), dtype=float32) Ny famindrana ny generators toy ny arguments ho an'ny Ny fandefasana Ny fandefasana Rehefa ampiasaina ho toy ny fanehoan-kevitra amin'ny a , ny zavatra samihafa amin'ny generator dia hiteraka ny famerenana ny . tf.function tf.function num_traces = 0 @tf.function def foo(g): global num_traces num_traces += 1 return g.normal([]) foo(tf.random.Generator.from_seed(1)) foo(tf.random.Generator.from_seed(2)) print(num_traces) 2 Ataovy azo antoka fa mifanaraka amin'ny : tf.Variable num_traces = 0 @tf.function def foo(v): global num_traces num_traces += 1 return v.read_value() foo(tf.Variable(1)) foo(tf.Variable(2)) print(num_traces) 1 Ny fifandraisana amin'ny sehatry ny fizarana Misy fomba roa izay Miara-miasa amin'ny sehatry ny fivarotana. Generator Ny famoronana generators ivelan'ny sehatry ny famoahana Raha ny generator dia naorina ivelan'ny sehatry ny politika, ny fidirana amin'ny replicas rehetra amin'ny generator dia ho serialy, ary noho izany ny replicas dia hahazo isan-karazany. g = tf.random.Generator.from_seed(1) strat = tf.distribute.MirroredStrategy(devices=["cpu:0", "cpu:1"]) with strat.scope(): def f(): print(g.normal([])) results = strat.run(f) INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') WARNING:tensorflow:Using MirroredStrategy eagerly has significant overhead currently. We will be working on improving this in the future, but for now please wrap `call_for_each_replica` or `experimental_run` or `run` inside a tf.function to get the best performance. tf.Tensor(0.43842274, shape=(), dtype=float32) tf.Tensor(1.6272374, shape=(), dtype=float32) Mariho fa ity fampiasana ity dia mety hanana olana amin'ny fampisehoana satria tsy mitovy amin'ny replicas ny fitaovana amin'ny generator. Ny famoronana generators ao anatin'ny sehatry ny famoahana Raha miorina amin'ny generator ao anatin'ny sehatry ny fitsipika, ny tsirairay dia hahazo rivotra samihafa sy tsy miankina amin'ny isan-karazany. strat = tf.distribute.MirroredStrategy(devices=["cpu:0", "cpu:1"]) with strat.scope(): g = tf.random.Generator.from_seed(1) print(strat.run(lambda: g.normal([]))) print(strat.run(lambda: g.normal([]))) INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') WARNING:tensorflow:Using MirroredStrategy eagerly has significant overhead currently. We will be working on improving this in the future, but for now please wrap `call_for_each_replica` or `experimental_run` or `run` inside a tf.function to get the best performance. PerReplica:{ 0: tf.Tensor(-0.87930447, shape=(), dtype=float32), 1: tf.Tensor(0.020661574, shape=(), dtype=float32) } WARNING:tensorflow:Using MirroredStrategy eagerly has significant overhead currently. We will be working on improving this in the future, but for now please wrap `call_for_each_replica` or `experimental_run` or `run` inside a tf.function to get the best performance. PerReplica:{ 0: tf.Tensor(-1.5822568, shape=(), dtype=float32), 1: tf.Tensor(0.77539235, shape=(), dtype=float32) } Fanamarihana: Amin'izao fotoana izao ny tf.random.Generator dia tsy manome safidy ahafahana mamela kopia samihafa hahazo (fa tsy mitovy) fantsona (izay tsy sarotra ara-teknika). Raha manana tranga fampiasanao ho an'ity endri-javatra ity ianao, azafady azafady azafady ny mpamorona TensorFlow. Fanamarihana: Amin'izao fotoana izao ny tf.random.Generator dia tsy manome safidy ahafahana mamela kopia samihafa hahazo (fa tsy mitovy) fantsona (izay tsy sarotra ara-teknika). Raha manana tranga fampiasanao ho an'ity endri-javatra ity ianao, azafady azafady azafady ny mpamorona TensorFlow. Raha ny mpamokatra dia voavonjy (ohatra, ny famoronana ), ny isa vahiny dia voafaritra amin'ny tsipika, na dia samy hafa replicas dia mahazo isan-karazany sy tsy mifandray isan-karazany. Azonao atao ny mieritreritra ny isa vahiny nateraky amin'ny replica toy ny hash ny replika ID sy ny "primary" isa vahiny izay mahazatra ho an'ny replika rehetra. Noho izany, ny rafitra manontolo dia mbola deterministika. Generator.from_seed Azonao atao ihany koa ny mamorona ao anatiny. : tf.random.Generator Strategy.run strat = tf.distribute.MirroredStrategy(devices=["cpu:0", "cpu:1"]) with strat.scope(): def f(): g = tf.random.Generator.from_seed(1) a = g.normal([]) b = g.normal([]) return tf.stack([a, b]) print(strat.run(f)) print(strat.run(f)) INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') WARNING:tensorflow:Using MirroredStrategy eagerly has significant overhead currently. We will be working on improving this in the future, but for now please wrap `call_for_each_replica` or `experimental_run` or `run` inside a tf.function to get the best performance. PerReplica:{ 0: tf.Tensor([-0.87930447 -1.5822568 ], shape=(2,), dtype=float32), 1: tf.Tensor([0.02066157 0.77539235], shape=(2,), dtype=float32) } WARNING:tensorflow:Using MirroredStrategy eagerly has significant overhead currently. We will be working on improving this in the future, but for now please wrap `call_for_each_replica` or `experimental_run` or `run` inside a tf.function to get the best performance. PerReplica:{ 0: tf.Tensor([-0.87930447 -1.5822568 ], shape=(2,), dtype=float32), 1: tf.Tensor([0.02066157 0.77539235], shape=(2,), dtype=float32) } Tsy manoro hevitra intsony ny handeha Araka ny hevitry ny Satria Amin'ny ankapobeny dia manantena fa ny arguments dia tendrors, fa tsy generators. tf.random.Generator Strategy.run Strategy.run Ny famonjena ny generator Amin'ny ankapobeny, ho an'ny famonjena na serialization dia azonao atao ny manatanteraka Amin'ny fomba mitovy amin'izany dia handray ny Ary a Ao amin'ny TF dia misy mekanisma roa ho an'ny serialization: Ary . tf.random.Generator tf.Variable tf.Module Ny fitsapana Ny modely Ny fitsapana Ny solosaina dia azo voavonjy sy hamerenana maimaim-poana amin'ny fampiasana Ny fivoaran'ny isan-karazany avy amin'ny toerana famerenana dia mitovy amin'ny avy amin'ny toerana fametrahana. tf.train.Checkpoint filename = "./checkpoint" g = tf.random.Generator.from_seed(1) cp = tf.train.Checkpoint(generator=g) print(g.normal([])) tf.Tensor(0.43842277, shape=(), dtype=float32) cp.write(filename) print("RNG stream from saving point:") print(g.normal([])) print(g.normal([])) RNG stream from saving point: tf.Tensor(1.6272374, shape=(), dtype=float32) tf.Tensor(1.6307176, shape=(), dtype=float32) cp.restore(filename) print("RNG stream from restoring point:") print(g.normal([])) print(g.normal([])) RNG stream from restoring point: tf.Tensor(1.6272374, shape=(), dtype=float32) tf.Tensor(1.6307176, shape=(), dtype=float32) Azonao atao ihany koa ny mamonjy sy hamerenana ao anatin'ny sehatry ny famoahana: filename = "./checkpoint" strat = tf.distribute.MirroredStrategy(devices=["cpu:0", "cpu:1"]) with strat.scope(): g = tf.random.Generator.from_seed(1) cp = tf.train.Checkpoint(my_generator=g) print(strat.run(lambda: g.normal([]))) INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') PerReplica:{ 0: tf.Tensor(-0.87930447, shape=(), dtype=float32), 1: tf.Tensor(0.020661574, shape=(), dtype=float32) } with strat.scope(): cp.write(filename) print("RNG stream from saving point:") print(strat.run(lambda: g.normal([]))) print(strat.run(lambda: g.normal([]))) RNG stream from saving point: PerReplica:{ 0: tf.Tensor(-1.5822568, shape=(), dtype=float32), 1: tf.Tensor(0.77539235, shape=(), dtype=float32) } PerReplica:{ 0: tf.Tensor(-0.5039703, shape=(), dtype=float32), 1: tf.Tensor(0.1251838, shape=(), dtype=float32) } with strat.scope(): cp.restore(filename) print("RNG stream from restoring point:") print(strat.run(lambda: g.normal([]))) print(strat.run(lambda: g.normal([]))) RNG stream from restoring point: PerReplica:{ 0: tf.Tensor(-1.5822568, shape=(), dtype=float32), 1: tf.Tensor(0.77539235, shape=(), dtype=float32) } PerReplica:{ 0: tf.Tensor(-0.5039703, shape=(), dtype=float32), 1: tf.Tensor(0.1251838, shape=(), dtype=float32) } Tokony ho azonao antoka fa tsy mitovy amin'ny tantaran'ny antso RNG ny replika (ohatra, ny iray amin'ireo replika dia manao antso RNG iray ary ny iray hafa manao antso RNG roa) alohan'ny fametrahana. (izay ihany no mamonjy ny toe-javatra amin'ny replika voalohany) dia tsy hamerenana ny replika rehetra araka ny tokony ho izy. tf.train.Checkpoint Azonao atao ihany koa ny mamerina ny fanaraha-maso voatahiry amin'ny sehatry ny famoahana hafa miaraka amin'ny isan'ny replika. Ny fitaovana natsangana tao amin'ny sehatra iray dia azo ampiasaina ihany koa amin'ny sehatra iray ihany, mba hamerina amin'ny sehatra hafa ianao dia tsy maintsy mamorona vaovao. Ao amin'ny sehatry ny tanjona sy ny vaovao Izany, araka ny hita ao amin'ity ohatra ity: tf.random.Generator tf.random.Generator tf.train.Checkpoint filename = "./checkpoint" strat1 = tf.distribute.MirroredStrategy(devices=["cpu:0", "cpu:1"]) with strat1.scope(): g1 = tf.random.Generator.from_seed(1) cp1 = tf.train.Checkpoint(my_generator=g1) print(strat1.run(lambda: g1.normal([]))) INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') PerReplica:{ 0: tf.Tensor(-0.87930447, shape=(), dtype=float32), 1: tf.Tensor(0.020661574, shape=(), dtype=float32) } with strat1.scope(): cp1.write(filename) print("RNG stream from saving point:") print(strat1.run(lambda: g1.normal([]))) print(strat1.run(lambda: g1.normal([]))) RNG stream from saving point: PerReplica:{ 0: tf.Tensor(-1.5822568, shape=(), dtype=float32), 1: tf.Tensor(0.77539235, shape=(), dtype=float32) } PerReplica:{ 0: tf.Tensor(-0.5039703, shape=(), dtype=float32), 1: tf.Tensor(0.1251838, shape=(), dtype=float32) } strat2 = tf.distribute.MirroredStrategy(devices=["cpu:0", "cpu:1", "cpu:2"]) with strat2.scope(): g2 = tf.random.Generator.from_seed(1) cp2 = tf.train.Checkpoint(my_generator=g2) cp2.restore(filename) print("RNG stream from restoring point:") print(strat2.run(lambda: g2.normal([]))) print(strat2.run(lambda: g2.normal([]))) INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1', '/job:localhost/replica:0/task:0/device:CPU:2') INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1', '/job:localhost/replica:0/task:0/device:CPU:2') RNG stream from restoring point: PerReplica:{ 0: tf.Tensor(-1.5822568, shape=(), dtype=float32), 1: tf.Tensor(0.77539235, shape=(), dtype=float32), 2: tf.Tensor(0.6851049, shape=(), dtype=float32) } PerReplica:{ 0: tf.Tensor(-0.5039703, shape=(), dtype=float32), 1: tf.Tensor(0.1251838, shape=(), dtype=float32), 2: tf.Tensor(-0.58519536, shape=(), dtype=float32) } Na dia Ary Ny fitaovana samihafa avy amin'ny Ary , dia mifandray amin'ny alàlan'ny rakitra Common Checkpoint Ny anaran'ny zavatra Mifandray amin'ny fomba fanao amin'ny fomba fanao (ohatra Ary Ity fiantohana ity dia tsy ahitana ny toe-javatra rehefa voatahiry ny generator ao amin'ny sehatry ny sehatry ny fitsipika ary naorina ivelan'ny sehatry ny fitsipika na mifanohitra amin'izany, satria ny fitaovana ivelan'ny sehatry ny fitsipika dia heverina ho hafa noho ny replika rehetra ao amin'ny fitsipika. g1 cp1 g2 cp2 filename my_generator cpu:0 cpu:1 Ny modely Mety ho voatahiry amin'ny SavedModel. Ny generator dia azo atao ao anatin'ny sehatry ny tetikady. Ny famonjena dia azo atao ihany koa ao anatin'ny sehatry ny tetikady. tf.random.Generator filename = "./saved_model" class MyModule(tf.Module): def __init__(self): super(MyModule, self).__init__() self.g = tf.random.Generator.from_seed(0) @tf.function def __call__(self): return self.g.normal([]) @tf.function def state(self): return self.g.state strat = tf.distribute.MirroredStrategy(devices=["cpu:0", "cpu:1"]) with strat.scope(): m = MyModule() print(strat.run(m)) print("state:", m.state()) INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:CPU:0', '/job:localhost/replica:0/task:0/device:CPU:1') PerReplica:{ 0: tf.Tensor(-1.4154755, shape=(), dtype=float32), 1: tf.Tensor(-0.11388441, shape=(), dtype=float32) } state: tf.Tensor([256 0 0], shape=(3,), dtype=int64) with strat.scope(): tf.saved_model.save(m, filename) print("RNG stream from saving point:") print(strat.run(m)) print("state:", m.state()) print(strat.run(m)) print("state:", m.state()) INFO:tensorflow:Assets written to: ./saved_model/assets INFO:tensorflow:Assets written to: ./saved_model/assets RNG stream from saving point: PerReplica:{ 0: tf.Tensor(-0.68758255, shape=(), dtype=float32), 1: tf.Tensor(0.8084062, shape=(), dtype=float32) } state: tf.Tensor([512 0 0], shape=(3,), dtype=int64) PerReplica:{ 0: tf.Tensor(-0.27342677, shape=(), dtype=float32), 1: tf.Tensor(-0.53093255, shape=(), dtype=float32) } state: tf.Tensor([768 0 0], shape=(3,), dtype=int64) imported = tf.saved_model.load(filename) print("RNG stream from loading point:") print("state:", imported.state()) print(imported()) print("state:", imported.state()) print(imported()) print("state:", imported.state()) RNG stream from loading point: state: tf.Tensor([256 0 0], shape=(3,), dtype=int64) tf.Tensor(-1.0359411, shape=(), dtype=float32) state: tf.Tensor([512 0 0], shape=(3,), dtype=int64) tf.Tensor(-0.06425078, shape=(), dtype=float32) state: tf.Tensor([768 0 0], shape=(3,), dtype=int64) Ampiasao ny SavedModel izay ahitana Ny fidirana amin'ny sehatry ny famoahana dia tsy ilaina satria ny replika rehetra dia hiteraka ny andaniny mitovy amin'ny isan-karazany (tsy izany satria ny ID replika dia voasokitra ao amin'ny sarin'i SavedModel). tf.random.Generator Ny fametrahana ny distributed Ny toe-javatra RNG dia hiverina araka ny tokony ho izy, fa ny isan'ny isan'ny voajanahary voajanahary dia tsy mitovy amin'ny voajanahary voalohany ao amin'ny rafitra (fa indray noho ny fitaovana ivelan'ny rafitra dia heverina tsy mitovy amin'ny replika rehetra ao amin'ny rafitra). tf.random.Generator Tsy misy ny RNGs Ny fampiasana RNGs tsy misy fanjakana dia tsotra. Satria izy ireo ihany no asa madio, dia tsy misy fiantraikany na fiantraikany eo amin'ny fanjakana. print(tf.random.stateless_normal(shape=[2, 3], seed=[1, 2])) print(tf.random.stateless_normal(shape=[2, 3], seed=[1, 2])) tf.Tensor( [[ 0.5441101 0.20738031 0.07356433] [ 0.04643455 -1.30159 -0.95385665]], shape=(2, 3), dtype=float32) tf.Tensor( [[ 0.5441101 0.20738031 0.07356433] [ 0.04643455 -1.30159 -0.95385665]], shape=(2, 3), dtype=float32) Ny RNG rehetra tsy misy fanjakana dia mitaky a arguments, izay tsy maintsy ho tanteraka Tensor ny endrika Ny vokatra amin'ny op dia voafaritra tanteraka amin'ity voa ity. seed [2] Ny algorithm RNG ampiasaina amin'ny RNG tsy misy fananana dia miankina amin'ny fitaovana, midika izany fa ny mitovy op mihazakazaka amin'ny fitaovana hafa dia mety hiteraka vokatra samihafa. Ny algorithm Ny jeneraly Ny roa Ny kilasy sy ny Ireo endri-javatra dia manohana ny algorithm Philox (amin'ny Ary Ao amin'ny fitaovana rehetra. tf.random.Generator stateless "philox" tf.random.Algorithm.PHILOX Ireo fitaovana samihafa dia hiteraka ny isan'ireo isan'ny tsirairay, raha mampiasa ny algorithm mitovy ary manomboka amin'ny toe-javatra mitovy izy ireo. dia hiteraka ihany koa ny isan'ny tsirairay, na dia mety hisy fahasamihafana kely eo amin'ny isan'ny fomba samihafa amin'ny fitaovana manatanteraka ny fanombanana ny tsirairay (ohatra ny fehezan-dalàna). Ny fitaovana XLA Ao amin'ny fitaovana miasa amin'ny XLA (ohatra TPU, ary koa CPU / GPU rehefa misy XLA) ny algorithm ThreeFry Ary Ity algorithm ity dia haingana amin'ny TPU fa lava amin'ny CPU / GPU raha oharina amin'ny Philox. "threefry" tf.random.Algorithm.THREEFRY Ny taratasy Ho an'ny antsipiriany bebe kokoa momba ireo algorithms. 'Parallel Random Numbers: Tsotra tahaka ny 1, 2, 3' Tamin'ny voalohany nivoaka tao amin'ny tranonkala TensorFlow, ity lahatsoratra ity dia hita eto amin'ny lohateny vaovao ary nahazo fahazoan-dalana amin'ny CC BY 4.0. Tamin'ny voalohany nivoaka tao amin'ny tranonkala TensorFlow, ity lahatsoratra ity dia hita eto amin'ny lohateny vaovao ary nahazo fahazoan-dalana amin'ny CC BY 4.0. Ny fifamoivoizana Ny fifamoivoizana