reknit guides

Author: t-kalinowski

man/layer_tfsm.Rd

@@ -213,7 +213,7 @@ layout_map["d1/bias"] <- tuple("model")
 layout_map["d2/output"] <- tuple("data", NULL)
 
 model_parallel <- keras$distribution$ModelParallel(
-  layout_map, batch_dim_name = "data"
+  layout_map = layout_map, batch_dim_name = "data"
 )
 
 keras$distribution$set_distribution(model_parallel)

vignettes-src/distribution.Rmd

@@ -967,7 +967,7 @@ model |> compile(
   optimizer = optimizer_rmsprop(1e-3),
   loss = list(
     loss_mean_squared_error(),
-    loss_categorical_crossentropy()
+    loss_binary_crossentropy()
   )
 )
 
@@ -986,6 +986,18 @@ model |> fit(
 )
 
 # Alternatively, fit on named lists (names matching)
+model <- keras_model(
+  inputs = list(image_input, timeseries_input),
+  outputs = list(score_output = score_output, 
+                 class_output = class_output)
+) |> compile(
+  optimizer = optimizer_rmsprop(1e-3),
+  loss = list(
+    loss_mean_squared_error(),
+    loss_binary_crossentropy()
+  )
+)
+
 model |> fit(
   list(img_input = img_data, ts_input = ts_data),
   list(score_output = score_targets, class_output = class_targets),

vignettes-src/training_with_built_in_methods.Rmd

@@ -127,11 +127,11 @@ model |> fit(x, y, epochs = 3)
 
 ```
 ## Epoch 1/3
-## 32/32 - 1s - 23ms/step - loss: 2.9118 - mae: 1.3597
+## 32/32 - 1s - 23ms/step - loss: 3.2271 - mae: 1.4339
 ## Epoch 2/3
-## 32/32 - 0s - 1ms/step - loss: 2.6026 - mae: 1.2856
+## 32/32 - 0s - 1ms/step - loss: 2.9034 - mae: 1.3605
 ## Epoch 3/3
-## 32/32 - 0s - 1ms/step - loss: 2.3378 - mae: 1.2193
+## 32/32 - 0s - 1ms/step - loss: 2.6272 - mae: 1.2960
 ```
 
 ## Going lower-level
@@ -212,11 +212,11 @@ model |> fit(x, y, epochs = 3)
 
 ```
 ## Epoch 1/3
-## 32/32 - 1s - 20ms/step - loss: 2.6540 - mae: 1.2901
+## 32/32 - 1s - 22ms/step - loss: 2.5170 - mae: 1.2923
 ## Epoch 2/3
-## 32/32 - 0s - 1ms/step - loss: 2.4139 - mae: 1.2303
+## 32/32 - 0s - 1ms/step - loss: 2.2689 - mae: 1.2241
 ## Epoch 3/3
-## 32/32 - 0s - 1ms/step - loss: 2.2080 - mae: 1.1761
+## 32/32 - 0s - 1ms/step - loss: 2.0578 - mae: 1.1633
 ```
 
 ## Supporting `sample_weight` & `class_weight`
@@ -282,11 +282,11 @@ model |> fit(x, y, sample_weight = sw, epochs = 3)
 
 ```
 ## Epoch 1/3
-## 32/32 - 1s - 25ms/step - loss: 0.1607 - mae: 1.3018
+## 32/32 - 1s - 26ms/step - loss: 0.1681 - mae: 1.3434
 ## Epoch 2/3
-## 32/32 - 0s - 1ms/step - loss: 0.1452 - mae: 1.2999
+## 32/32 - 0s - 9ms/step - loss: 0.1394 - mae: 1.3364
 ## Epoch 3/3
-## 32/32 - 0s - 1ms/step - loss: 0.1335 - mae: 1.2986
+## 32/32 - 0s - 1ms/step - loss: 0.1148 - mae: 1.3286
 ```
 
 ## Providing your own evaluation step
@@ -332,15 +332,15 @@ model |> evaluate(x, y)
 ```
 
 ```
-## 32/32 - 0s - 9ms/step - loss: 0.0000e+00 - mae: 1.3947
+## 32/32 - 0s - 10ms/step - loss: 0.0000e+00 - mae: 1.3871
 ```
 
 ```
 ## $loss
 ## [1] 0
 ##
 ## $mae
-## [1] 1.394695
+## [1] 1.387149
 ```
 
 ## Wrapping up: an end-to-end GAN example
@@ -507,7 +507,7 @@ gan |> fit(
 ```
 
 ```
-## 100/100 - 5s - 51ms/step - d_loss: 0.0000e+00 - g_loss: 0.0000e+00
+## 100/100 - 5s - 54ms/step - d_loss: 0.0000e+00 - g_loss: 0.0000e+00
 ```
 
 The ideas behind deep learning are simple, so why should their implementation be painful?

vignettes/custom_train_step_in_tensorflow.Rmd

@@ -193,18 +193,18 @@ with(strategy$scope(), {
 
 ```
 ## Epoch 1/2
-## 782/782 - 7s - 9ms/step - loss: 3.0622 - sparse_categorical_accuracy: 0.8615 - val_loss: 1.1367 - val_sparse_categorical_accuracy: 0.9006
+## 782/782 - 4s - 6ms/step - loss: 2.1409 - sparse_categorical_accuracy: 0.8896 - val_loss: 0.7223 - val_sparse_categorical_accuracy: 0.9216
 ## Epoch 2/2
-## 782/782 - 4s - 5ms/step - loss: 0.5774 - sparse_categorical_accuracy: 0.9259 - val_loss: 0.6612 - val_sparse_categorical_accuracy: 0.9210
-## 157/157 - 1s - 3ms/step - loss: 0.6729 - sparse_categorical_accuracy: 0.9150
+## 782/782 - 3s - 4ms/step - loss: 0.4292 - sparse_categorical_accuracy: 0.9387 - val_loss: 0.3693 - val_sparse_categorical_accuracy: 0.9404
+## 157/157 - 0s - 2ms/step - loss: 0.3976 - sparse_categorical_accuracy: 0.9386
 ```
 
 ```
 ## $loss
-## [1] 0.6728871
+## [1] 0.3976028
 ##
 ## $sparse_categorical_accuracy
-## [1] 0.915
+## [1] 0.9386
 ```
 
 ## Using callbacks to ensure fault tolerance
@@ -274,7 +274,7 @@ run_training(epochs = 1)
 ```
 
 ```
-## 782/782 - 3s - 4ms/step - loss: 0.2194 - sparse_categorical_accuracy: 0.9532 - val_loss: 0.3217 - val_sparse_categorical_accuracy: 0.9456
+## 782/782 - 4s - 5ms/step - loss: 0.1485 - sparse_categorical_accuracy: 0.9627 - val_loss: 0.2062 - val_sparse_categorical_accuracy: 0.9560
 ```
 
 ``` r
@@ -283,7 +283,7 @@ run_training(epochs = 1)
 ```
 
 ```
-## 782/782 - 3s - 4ms/step - loss: 0.1876 - sparse_categorical_accuracy: 0.9584 - val_loss: 0.3629 - val_sparse_categorical_accuracy: 0.9396
+## 782/782 - 4s - 5ms/step - loss: 0.1227 - sparse_categorical_accuracy: 0.9673 - val_loss: 0.2007 - val_sparse_categorical_accuracy: 0.9602
 ```
 
 ## `tf$data` performance tips

vignettes/distributed_training_with_tensorflow.Rmd

@@ -184,24 +184,24 @@ model |> fit(dataset, epochs = 3)
 
 ```
 ## Epoch 1/3
-## 8/8 - 0s - 37ms/step - loss: 1.0629
+## 8/8 - 0s - 38ms/step - loss: 1.0768
 ## Epoch 2/3
-## 8/8 - 0s - 4ms/step - loss: 0.9712
+## 8/8 - 0s - 6ms/step - loss: 0.9754
 ## Epoch 3/3
-## 8/8 - 0s - 5ms/step - loss: 0.9322
+## 8/8 - 0s - 5ms/step - loss: 0.9347
 ```
 
 ``` r
 model |> evaluate(dataset)
 ```
 
 ```
-## 8/8 - 0s - 7ms/step - loss: 0.8859
+## 8/8 - 0s - 7ms/step - loss: 0.8936
 ```
 
 ```
 ## $loss
-## [1] 0.8858577
+## [1] 0.8935966
 ```
 
 
@@ -253,8 +253,7 @@ layout_map["d1/bias"] <- tuple("model")
 layout_map["d2/output"] <- tuple("data", NULL)
 
 model_parallel <- keras$distribution$ModelParallel(
-  layout_map = layout_map,
-  batch_dim_name = "data"
+  layout_map = layout_map, batch_dim_name = "data"
 )
 
 keras$distribution$set_distribution(model_parallel)
@@ -279,24 +278,24 @@ model |> fit(dataset, epochs = 3)
 
 ```
 ## Epoch 1/3
-## 8/8 - 0s - 29ms/step - loss: 1.0714
+## 8/8 - 0s - 29ms/step - loss: 1.0836
 ## Epoch 2/3
-## 8/8 - 0s - 4ms/step - loss: 0.9744
+## 8/8 - 0s - 4ms/step - loss: 1.0192
 ## Epoch 3/3
-## 8/8 - 0s - 3ms/step - loss: 0.9280
+## 8/8 - 0s - 4ms/step - loss: 0.9821
 ```
 
 ``` r
 model |> evaluate(dataset)
 ```
 
 ```
-## 8/8 - 0s - 9ms/step - loss: 0.8802
+## 8/8 - 0s - 8ms/step - loss: 0.9576
 ```
 
 ```
 ## $loss
-## [1] 0.8802156
+## [1] 0.9576273
 ```
 
 

vignettes/distribution.Rmd

@@ -208,17 +208,17 @@ history <- model |> fit(
 
 ```
 ## Epoch 1/2
-## 750/750 - 2s - 2ms/step - accuracy: 0.8979 - loss: 0.3540 - val_accuracy: 0.9448 - val_loss: 0.1903
+## 750/750 - 2s - 3ms/step - accuracy: 0.8979 - loss: 0.3540 - val_accuracy: 0.9448 - val_loss: 0.1903
 ## Epoch 2/2
-## 750/750 - 1s - 801us/step - accuracy: 0.9511 - loss: 0.1634 - val_accuracy: 0.9605 - val_loss: 0.1386
+## 750/750 - 1s - 784us/step - accuracy: 0.9509 - loss: 0.1635 - val_accuracy: 0.9597 - val_loss: 0.1397
 ```
 
 ``` r
 test_scores <- model |> evaluate(x_test, y_test, verbose=2)
 ```
 
 ```
-## 313/313 - 0s - 1ms/step - accuracy: 0.9593 - loss: 0.1323
+## 313/313 - 0s - 1ms/step - accuracy: 0.9595 - loss: 0.1328
 ```
 
 
@@ -229,8 +229,8 @@ cat("Test accuracy:", test_scores$accuracy, "\n")
 ```
 
 ```
-## Test loss: 0.1323339
-## Test accuracy: 0.9593
+## Test loss: 0.132778
+## Test accuracy: 0.9595
 ```
 
 For further reading, see the [training and evaluation](training_with_built_in_methods.html) guide.
@@ -643,9 +643,9 @@ model |> fit(
 
 ```
 ## Epoch 1/2
-## 40/40 - 2s - 62ms/step - loss: 348.4560
+## 40/40 - 3s - 66ms/step - department_loss: 381.0319 - loss: 498.7886 - priority_loss: 117.7568
 ## Epoch 2/2
-## 40/40 - 0s - 6ms/step - loss: 255.0283
+## 40/40 - 0s - 10ms/step - department_loss: 358.5705 - loss: 438.4252 - priority_loss: 79.8546
 ```
 
 When calling fit with a `Dataset` object, it should yield either a
@@ -776,7 +776,7 @@ model |> fit(
 ```
 
 ```
-## 13/13 - 5s - 364ms/step - acc: 0.1312 - loss: 2.3010 - val_acc: 0.1200 - val_loss: 2.2964
+## 13/13 - 5s - 375ms/step - acc: 0.1250 - loss: 2.3001 - val_acc: 0.1400 - val_loss: 2.2938
 ```
 
 ## Shared layers

vignettes/functional_api.Rmd

@@ -196,15 +196,15 @@ model |> evaluate(x_test, y_test)
 ```
 
 ```
-## 313/313 - 1s - 3ms/step - accuracy: 0.9806 - loss: 0.0880
+## 313/313 - 1s - 2ms/step - accuracy: 0.9816 - loss: 0.0847
 ```
 
 ```
 ## $accuracy
-## [1] 0.9806
+## [1] 0.9816
 ##
 ## $loss
-## [1] 0.08797418
+## [1] 0.08465149
 ```
 
 Generate predictions on new data:
@@ -223,7 +223,7 @@ max.col(probs) - 1L
 ```
 
 ```
-##   [1] 7 2 1 0 4 1 4 9 6 9 0 6 9 0 1 5 9 7 3 4 9 6 6 5 4 0 7 4 0 1 3 1 3 4 7
+##   [1] 7 2 1 0 4 1 4 9 5 9 0 6 9 0 1 5 9 7 3 4 9 6 6 5 4 0 7 4 0 1 3 1 3 4 7
 ##  [36] 2 7 1 2 1 1 7 4 2 3 5 1 2 4 4 6 3 5 5 6 0 4 1 9 5 7 8 9 3 7 4 6 4 3 0
 ##  [71] 7 0 2 9 1 7 3 2 9 7 7 6 2 7 8 4 7 3 6 1 3 6 9 3 1 4 1 7 6 9
 ##  [ reached getOption("max.print") -- omitted 9900 entries ]

vignettes/getting_started.Rmd

@@ -33,9 +33,6 @@ edit the string below to `"jax"` or `"torch"` and hit
 This entire guide is backend-agnostic.
 
 
-
-
-
 ``` r
 library(tensorflow, exclude = c("shape", "set_random_seed"))
 library(keras3)
@@ -181,25 +178,25 @@ model |> fit(
 
 ```
 ## Epoch 1/10
-## 399/399 - 7s - 16ms/step - acc: 0.7495 - loss: 0.7390 - val_acc: 0.9644 - val_loss: 0.1219
+## 399/399 - 8s - 20ms/step - acc: 0.7476 - loss: 0.7467 - val_acc: 0.9663 - val_loss: 0.1179
 ## Epoch 2/10
-## 399/399 - 2s - 5ms/step - acc: 0.9384 - loss: 0.2051 - val_acc: 0.9758 - val_loss: 0.0794
+## 399/399 - 2s - 5ms/step - acc: 0.9384 - loss: 0.2066 - val_acc: 0.9770 - val_loss: 0.0765
 ## Epoch 3/10
-## 399/399 - 2s - 5ms/step - acc: 0.9567 - loss: 0.1468 - val_acc: 0.9809 - val_loss: 0.0632
+## 399/399 - 2s - 5ms/step - acc: 0.9569 - loss: 0.1467 - val_acc: 0.9817 - val_loss: 0.0622
 ## Epoch 4/10
-## 399/399 - 2s - 5ms/step - acc: 0.9656 - loss: 0.1167 - val_acc: 0.9857 - val_loss: 0.0479
+## 399/399 - 2s - 5ms/step - acc: 0.9652 - loss: 0.1170 - val_acc: 0.9860 - val_loss: 0.0499
 ## Epoch 5/10
-## 399/399 - 2s - 5ms/step - acc: 0.9716 - loss: 0.0984 - val_acc: 0.9883 - val_loss: 0.0427
+## 399/399 - 2s - 5ms/step - acc: 0.9709 - loss: 0.0999 - val_acc: 0.9873 - val_loss: 0.0447
 ## Epoch 6/10
-## 399/399 - 2s - 5ms/step - acc: 0.9756 - loss: 0.0852 - val_acc: 0.9879 - val_loss: 0.0412
+## 399/399 - 2s - 5ms/step - acc: 0.9752 - loss: 0.0863 - val_acc: 0.9877 - val_loss: 0.0400
 ## Epoch 7/10
-## 399/399 - 2s - 5ms/step - acc: 0.9765 - loss: 0.0786 - val_acc: 0.9894 - val_loss: 0.0394
+## 399/399 - 2s - 5ms/step - acc: 0.9764 - loss: 0.0787 - val_acc: 0.9890 - val_loss: 0.0395
 ## Epoch 8/10
-## 399/399 - 2s - 5ms/step - acc: 0.9794 - loss: 0.0672 - val_acc: 0.9884 - val_loss: 0.0415
+## 399/399 - 2s - 5ms/step - acc: 0.9794 - loss: 0.0678 - val_acc: 0.9874 - val_loss: 0.0432
 ## Epoch 9/10
-## 399/399 - 2s - 5ms/step - acc: 0.9808 - loss: 0.0647 - val_acc: 0.9901 - val_loss: 0.0369
+## 399/399 - 2s - 5ms/step - acc: 0.9802 - loss: 0.0658 - val_acc: 0.9894 - val_loss: 0.0395
 ## Epoch 10/10
-## 399/399 - 2s - 5ms/step - acc: 0.9836 - loss: 0.0571 - val_acc: 0.9911 - val_loss: 0.0325
+## 399/399 - 2s - 5ms/step - acc: 0.9825 - loss: 0.0584 - val_acc: 0.9914 - val_loss: 0.0342
 ```
 
 ``` r
@@ -230,7 +227,7 @@ predictions <- model |> predict(x_test)
 ```
 
 ```
-## 313/313 - 1s - 2ms/step
+## 313/313 - 0s - 2ms/step
 ```
 
 ``` r
@@ -365,7 +362,7 @@ model |> fit(
 ```
 
 ```
-## 399/399 - 6s - 15ms/step - acc: 0.7355 - loss: 0.7722 - val_acc: 0.9272 - val_loss: 0.2380
+## 399/399 - 6s - 15ms/step - acc: 0.7343 - loss: 0.7741 - val_acc: 0.9269 - val_loss: 0.2399
 ```
 
 ## Training models on arbitrary data sources
@@ -442,7 +439,7 @@ model |> fit(train_dataset, epochs = 1, validation_data = test_dataset)
 ```
 
 ```
-## 469/469 - 7s - 15ms/step - acc: 0.7492 - loss: 0.7481 - val_acc: 0.9112 - val_loss: 0.3002
+## 469/469 - 7s - 14ms/step - acc: 0.7499 - loss: 0.7454 - val_acc: 0.9051 - val_loss: 0.3089
 ```
 
 ## Further reading