update vignett tethers

Author: t-kalinowski

.tether/vignettes-src/parked/_distributed_training_with_jax.Rmd

@@ -174,7 +174,7 @@ optimizer.build(model.trainable_variables)
 # Keras provides a pure functional forward pass: model.stateless_call
 def compute_loss(trainable_variables, non_trainable_variables, x, y):
     y_pred, updated_non_trainable_variables = model.stateless_call(
-        trainable_variables, non_trainable_variables, x
+        trainable_variables, non_trainable_variables, x, training=True
     )
     loss_value = loss(y, y_pred)
     return loss_value, updated_non_trainable_variables

.tether/vignettes-src/parked/_writing_a_custom_training_loop_in_jax.Rmd

@@ -175,7 +175,7 @@ variables.
 ```python
 def compute_loss_and_updates(trainable_variables, non_trainable_variables, x, y):
     y_pred, non_trainable_variables = model.stateless_call(
-        trainable_variables, non_trainable_variables, x
+        trainable_variables, non_trainable_variables, x, training=True
     )
     loss = loss_fn(y, y_pred)
     return loss, non_trainable_variables

tools/retether.R

@@ -47,8 +47,8 @@ resolve_rmd_tether <- function(url) {
   tutobook_to_rmd(path, outfile = FALSE)
 }
 
-resolve_rmd_tether <- NULL
-# resolve_roxy_tether <- NULL
+# resolve_rmd_tether <- NULL
+resolve_roxy_tether <- NULL
 
 
 # options(warn = 2)

vignettes-src/parked/_distributed_training_with_jax.Rmd

@@ -174,7 +174,7 @@ optimizer.build(model.trainable_variables)
 # Keras provides a pure functional forward pass: model.stateless_call
 def compute_loss(trainable_variables, non_trainable_variables, x, y):
     y_pred, updated_non_trainable_variables = model.stateless_call(
-        trainable_variables, non_trainable_variables, x
+        trainable_variables, non_trainable_variables, x, training=True
     )
     loss_value = loss(y, y_pred)
     return loss_value, updated_non_trainable_variables

vignettes-src/parked/_writing_a_custom_training_loop_in_jax.Rmd

@@ -175,7 +175,7 @@ variables.
 ```python
 def compute_loss_and_updates(trainable_variables, non_trainable_variables, x, y):
     y_pred, non_trainable_variables = model.stateless_call(
-        trainable_variables, non_trainable_variables, x
+        trainable_variables, non_trainable_variables, x, training=True
     )
     loss = loss_fn(y, y_pred)
     return loss, non_trainable_variables