[YDF] [TF-DF] Minor fixes

- Pass release_cpu_linux flag to test_bazel to on linux for manylinux compatibility
- Remove the MacOS Tensorflow warning
- Fix an incorrect warning on max_depth =-1 or -2, which are special cases (and always set)

PiperOrigin-RevId: 665904417

yggdrasil_decision_forests/learner/decision_tree/generic_parameters.cc

@@ -460,7 +460,8 @@ absl::Status SetHyperParameters(
     const auto hparam = generic_hyper_params->Get(kHParamMaxDepth);
     if (hparam.has_value()) {
       dt_config->set_max_depth(hparam.value().value().integer());
-      if (dt_config->max_depth() < 2) {
+      if (dt_config->max_depth() < 2 && dt_config->max_depth() != -1 &&
+          dt_config->max_depth() != -2) {
         LOG(WARNING)
             << "Setting max_depth=" << dt_config->max_depth()
             << " for a training model will not result in any learning (i.e. "

yggdrasil_decision_forests/port/python/ydf/model/generic_model.py

@@ -984,18 +984,11 @@ def to_tensorflow_function(  # pytype: disable=name-error
         classification, outputs a tensorflow of shape [num examples, 2]
         containing the probability of both the negative and positive classes.
         Has no effect on non-binary classification models.
-      force: Try to export even in currently unsupported environments. WARNING:
-        Setting this to true may crash the Python runtime.
+      force: Try to export even in currently unsupported environments.
 
     Returns:
       A TensorFlow @tf.function.
     """
-    if platform.system() == "Darwin" and not force:
-      raise ValueError(
-          "Exporting to TensorFlow is currently broken on MacOS and may crash"
-          " the current Python process. To proceed anyway, add parameter"
-          " `force=True`."
-      )
 
     return _get_export_tf().ydf_model_to_tf_function(
         ydf_model=self,