Mengapa GradientTape Adalah Ciri yang Paling Disingkirkan dalam TensorFlow

Gambaran keseluruhan kandungan Perbezaan automatik dan gradient Setup Gradient Komputer Tape Gradient Gradient berbanding model Mengawal apa yang menonton pita Hasil pertengahan Nota mengenai prestasi Gradien matlamat bukan skalar Kawalan aliran Perbezaan automatik dan gradient berguna untuk melaksanakan algoritma pembelajaran mesin seperti Melatih rangkaian saraf. Perbezaan automatik Backpropagasi Dalam panduan ini, anda akan meneroka cara untuk mengira gradient dengan TensorFlow, terutamanya dalam pelaksanaan yang bersemangat. Setup import numpy as np import matplotlib.pyplot as plt import tensorflow as tf 2024-08-15 01:30:07.003169: 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 01:30:07.023862: 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 01:30:07.029954: 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 Gradient Komputer Untuk membezakan secara automatik, TensorFlow perlu mengingati operasi apa yang berlaku dalam urutan apa semasa Selepas itu, semasa , TensorFlow melintasi senarai operasi ini dalam urutan terbalik untuk mengira gradien. ke hadapan Kembali Pass Tape Gradient TensorFlow menyediakan API untuk diferensiasi automatik; iaitu, mengira gradien perhitungan sehubungan dengan beberapa input, biasanya s. TensorFlow "mencatat" operasi yang berkaitan yang dijalankan dalam konteks TensorFlow kemudian menggunakan pita itu untuk mengira gradien pengiraan "diterima" menggunakan . tf.GradientTape tf.Variable tf.GradientTape Cara membalikkan perbezaan Berikut ialah contoh mudah: x = tf.Variable(3.0) with tf.GradientTape() as tape: y = x**2 WARNING: All log messages before absl::InitializeLog() is called are written to STDERR I0000 00:00:1723685409.408818 20970 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 00:00:1723685409.412555 20970 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 00:00:1723685409.416343 20970 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 00:00:1723685409.420087 20970 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 00:00:1723685409.431667 20970 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 00:00:1723685409.435229 20970 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 00:00:1723685409.438777 20970 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 00:00:1723685409.442350 20970 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 00:00:1723685409.445712 20970 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 00:00:1723685409.449141 20970 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 00:00:1723685409.452491 20970 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 00:00:1723685409.456034 20970 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 00:00:1723685410.685265 20970 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 00:00:1723685410.687389 20970 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 00:00:1723685410.689411 20970 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 00:00:1723685410.691490 20970 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 00:00:1723685410.693542 20970 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 00:00:1723685410.695541 20970 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 00:00:1723685410.697441 20970 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 00:00:1723685410.699432 20970 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 00:00:1723685410.701351 20970 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 00:00:1723685410.703333 20970 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 00:00:1723685410.705229 20970 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 00:00:1723685410.707222 20970 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 00:00:1723685410.744994 20970 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 00:00:1723685410.747037 20970 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 00:00:1723685410.749507 20970 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 00:00:1723685410.751538 20970 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 00:00:1723685410.753500 20970 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 00:00:1723685410.755501 20970 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 00:00:1723685410.757421 20970 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 00:00:1723685410.759404 20970 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 00:00:1723685410.761363 20970 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 00:00:1723685410.763858 20970 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 00:00:1723685410.766199 20970 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 00:00:1723685410.768560 20970 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 Sebaik sahaja anda telah merekodkan beberapa operasi, gunakan untuk mengira gradien beberapa sasaran (seringkali kerugian) berbanding dengan beberapa sumber (seringkali variabel model): GradientTape.gradient(target, sources) # dy = 2x * dx dy_dx = tape.gradient(y, x) dy_dx.numpy() 6.0 Contoh di atas menggunakan skalar, tetapi berfungsi dengan mudah pada mana-mana tensor: tf.GradientTape w = tf.Variable(tf.random.normal((3, 2)), name='w') b = tf.Variable(tf.zeros(2, dtype=tf.float32), name='b') x = [[1., 2., 3.]] with tf.GradientTape(persistent=True) as tape: y = x @ w + b loss = tf.reduce_mean(y**2) Untuk mendapatkan gradient daripada berkenaan dengan kedua-dua variabel, anda boleh melepasi kedua-dua sebagai sumber kepada Pita ini fleksibel tentang bagaimana sumber dihantar dan akan menerima mana-mana gabungan senarai atau kamus dan mengembalikan gradient berstruktur dengan cara yang sama (lihat ) daripada loss gradient tf.nest [dl_dw, dl_db] = tape.gradient(loss, [w, b]) Gradient berkenaan dengan setiap sumber mempunyai bentuk sumber: print(w.shape) print(dl_dw.shape) (3, 2) (3, 2) Berikut adalah pengiraan gradient lagi, kali ini melepasi kamus variabel: my_vars = { 'w': w, 'b': b } grad = tape.gradient(loss, my_vars) grad['b'] Gradient berbanding model Ia adalah biasa untuk mengumpul dalam a atau salah satu daripada subkelasnya ( daripada ) untuk dan . tf.Variables tf.Module layers.Layer keras.Model Pengesahan Pengeksportan Dalam kebanyakan kes, anda akan mahu mengira gradient sehubungan dengan variabel yang boleh dilatih model. mengumpulkan variabel-variabel mereka dalam properti, anda boleh mengira gradien ini dalam beberapa baris kod: tf.Module Module.trainable_variables layer = tf.keras.layers.Dense(2, activation='relu') x = tf.constant([[1., 2., 3.]]) with tf.GradientTape() as tape: # Forward pass y = layer(x) loss = tf.reduce_mean(y**2) # Calculate gradients with respect to every trainable variable grad = tape.gradient(loss, layer.trainable_variables) for var, g in zip(layer.trainable_variables, grad): print(f'{var.name}, shape: {g.shape}') kernel, shape: (3, 2) bias, shape: (2,) Mengawal apa yang menonton pita Tindakan lalai ialah untuk merekodkan semua operasi selepas mengakses Sebab-sebab untuk ini ialah: tf.Variable Pita perlu tahu operasi mana yang perlu disenaraikan dalam laluan ke hadapan untuk mengira gradien dalam laluan ke hadapan. Pita mengandungi rujukan kepada output pertengahan, jadi anda tidak mahu merekodkan operasi yang tidak perlu. Kes penggunaan yang paling biasa melibatkan mengira gradient kerugian sehubungan dengan semua variabel yang boleh dilatih dalam model. Contohnya, yang berikut tidak boleh mengira gradien kerana tidak boleh "menonton" secara lalai, dan Tidak boleh dilatih: tf.Tensor tf.Variable # A trainable variable x0 = tf.Variable(3.0, name='x0') # Not trainable x1 = tf.Variable(3.0, name='x1', trainable=False) # Not a Variable: A variable + tensor returns a tensor. x2 = tf.Variable(2.0, name='x2') + 1.0 # Not a variable x3 = tf.constant(3.0, name='x3') with tf.GradientTape() as tape: y = (x0**2) + (x1**2) + (x2**2) grad = tape.gradient(y, [x0, x1, x2, x3]) for g in grad: print(g) tf.Tensor(6.0, shape=(), dtype=float32) None None None Anda boleh senaraikan variabel yang sedang dilihat oleh pita menggunakan kaedah : GradientTape.watched_variables [var.name for var in tape.watched_variables()] ['x0:0'] menyediakan hooks yang memberi pengguna kawalan atas apa yang sedang atau tidak dilihat. tf.GradientTape Untuk mencatatkan gradient berbanding a Anda perlu memanggil : tf.Tensor GradientTape.watch(x) x = tf.constant(3.0) with tf.GradientTape() as tape: tape.watch(x) y = x**2 # dy = 2x * dx dy_dx = tape.gradient(y, x) print(dy_dx.numpy()) 6.0 Sebaliknya, untuk mematikan tingkah laku lalai menonton semua • Set Perhitungan ini menggunakan dua variabel, tetapi hanya menghubungkan gradient untuk salah satu variabel: tf.Variables watch_accessed_variables=False x0 = tf.Variable(0.0) x1 = tf.Variable(10.0) with tf.GradientTape(watch_accessed_variables=False) as tape: tape.watch(x1) y0 = tf.math.sin(x0) y1 = tf.nn.softplus(x1) y = y0 + y1 ys = tf.reduce_sum(y) Sejak Tidak dipanggil kepada , tidak ada gradient dikira berkaitan dengannya: GradientTape.watch x0 # dys/dx1 = exp(x1) / (1 + exp(x1)) = sigmoid(x1) grad = tape.gradient(ys, {'x0': x0, 'x1': x1}) print('dy/dx0:', grad['x0']) print('dy/dx1:', grad['x1'].numpy()) dy/dx0: None dy/dx1: 0.9999546 Hasil pertengahan Anda juga boleh meminta gradien output sehubungan dengan nilai pertengahan yang dikira dalam dalam konteks tf.GradientTape x = tf.constant(3.0) with tf.GradientTape() as tape: tape.watch(x) y = x * x z = y * y # Use the tape to compute the gradient of z with respect to the # intermediate value y. # dz_dy = 2 * y and y = x ** 2 = 9 print(tape.gradient(z, y).numpy()) 18.0 Secara lalai, sumber-sumber yang dimiliki oleh a Mereka akan dibebaskan sebaik sahaja Untuk mengira beberapa gradient di atas pengiraan yang sama, cipta pita gradient dengan Ia membolehkan pelbagai panggilan kepada kaedah sebagai sumber dibebaskan apabila objek pita adalah sampah dikumpulkan. GradientTape GradientTape.gradient persistent=True gradient x = tf.constant([1, 3.0]) with tf.GradientTape(persistent=True) as tape: tape.watch(x) y = x * x z = y * y print(tape.gradient(z, x).numpy()) # [4.0, 108.0] (4 * x**3 at x = [1.0, 3.0]) print(tape.gradient(y, x).numpy()) # [2.0, 6.0] (2 * x at x = [1.0, 3.0]) [ 4. 108.] [2. 6.] del tape # Drop the reference to the tape Nota mengenai prestasi There is a tiny overhead associated with doing operations inside a gradient tape context. For most eager execution this will not be a noticeable cost, but you should still use tape context around the areas only where it is required. Gradient tapes use memory to store intermediate results, including inputs and outputs, for use during the backwards pass. For efficiency, some ops (like ) don't need to keep their intermediate results and they are pruned during the forward pass. However, if you use on your tape, and your peak memory usage will be higher. ReLU persistent=True nothing is discarded Gradien matlamat bukan skalar Gradient ialah operasi pada skalar. x = tf.Variable(2.0) with tf.GradientTape(persistent=True) as tape: y0 = x**2 y1 = 1 / x print(tape.gradient(y0, x).numpy()) print(tape.gradient(y1, x).numpy()) 4.0 -0.25 Oleh itu, jika anda meminta untuk gradient pelbagai sasaran, hasil untuk setiap sumber adalah: Gradient daripada jumlah sasaran, atau setara Jumlah gradient setiap sasaran. x = tf.Variable(2.0) with tf.GradientTape() as tape: y0 = x**2 y1 = 1 / x print(tape.gradient({'y0': y0, 'y1': y1}, x).numpy()) 3.75 Begitu juga, jika sasaran (s) tidak skalar, gradient jumlah dikira: x = tf.Variable(2.) with tf.GradientTape() as tape: y = x * [3., 4.] print(tape.gradient(y, x).numpy()) 7.0 Ini menjadikannya mudah untuk mengambil gradient jumlah koleksi kerugian, atau gradient jumlah pengiraan kerugian elemen. Jika anda memerlukan gradient berasingan untuk setiap item, rujuk kepada . Jacobian Dalam sesetengah kes, anda boleh melepaskan Jacobian.Untuk pengiraan gaya elemen, gradien jumlah memberikan derivatif setiap elemen sehubungan dengan input-elemennya, kerana setiap elemen adalah bebas: x = tf.linspace(-10.0, 10.0, 200+1) with tf.GradientTape() as tape: tape.watch(x) y = tf.nn.sigmoid(x) dy_dx = tape.gradient(y, x) plt.plot(x, y, label='y') plt.plot(x, dy_dx, label='dy/dx') plt.legend() _ = plt.xlabel('x') Kawalan aliran Kerana pita gradient mencatat operasi apabila mereka dijalankan, aliran kawalan Python diproses secara semulajadi (contohnya, dan daripada pernyataan). if while Di sini, satu variabel yang berbeza digunakan pada setiap cawangan Gradient hanya menyambung kepada variabel yang digunakan: if x = tf.constant(1.0) v0 = tf.Variable(2.0) v1 = tf.Variable(2.0) with tf.GradientTape(persistent=True) as tape: tape.watch(x) if x > 0.0: result = v0 else: result = v1**2 dv0, dv1 = tape.gradient(result, [v0, v1]) print(dv0) print(dv1) tf.Tensor(1.0, shape=(), dtype=float32) None Hanya ingat bahawa pernyataan kawalan itu sendiri tidak boleh dibezakan, jadi mereka tidak kelihatan kepada pengoptimum berasaskan gradient. Bergantung kepada nilai dalam contoh di atas, rakaman atau rekod atau Gradient berkenaan dengan sentiasa . x result = v0 result = v1**2 x None dx = tape.gradient(result, x) print(dx) None Awalnya diterbitkan di laman web TensorFlow, artikel ini muncul di sini di bawah tajuk baru dan dilesenkan di bawah CC BY 4.0. Awalnya diterbitkan di laman web TensorFlow, artikel ini muncul di sini di bawah tajuk baru dan dilesenkan di bawah CC BY 4.0. TensorFlow TensorFlow