Fix Problem.filepattern to include PREDICT

Ryan Sepassi · Ryan Sepassi · commit 3950b4027ac5 · 2017-09-29T14:59:47.000-07:00
PiperOrigin-RevId: 170415717
diff --git a/tensor2tensor/data_generators/problem.py b/tensor2tensor/data_generators/problem.py
@@ -235,12 +235,26 @@ def test_filepaths(self, data_dir, num_shards, shuffled):
                                                num_shards)
 
   def filepattern(self, data_dir, mode):
-    """Get filepattern for data files for mode."""
+    """Get filepattern for data files for mode.
+
+    Matches mode to a suffix.
+    * TRAIN: train
+    * EVAL: dev
+    * PREDICT: dev
+    * test: test
+
+    Args:
+      data_dir: str, data directory.
+      mode: tf.estimator.ModeKeys or "test".
+
+    Returns:
+      filepattern str
+    """
     path = os.path.join(data_dir, self.dataset_filename())
 
     if mode == tf.estimator.ModeKeys.TRAIN:
       suffix = "train"
-    elif mode == tf.estimator.ModeKeys.EVAL:
+    elif mode in [tf.estimator.ModeKeys.EVAL, tf.estimator.ModeKeys.PREDICT]:
       suffix = "dev"
     else:
       assert mode == "test"
diff --git a/tensor2tensor/utils/model_builder.py b/tensor2tensor/utils/model_builder.py
@@ -288,7 +288,7 @@ def nth_model(n):
   diet_vars = [
       v for v in tf.global_variables() if v.dtype == dtypes.float16_ref
   ]
-  _log_variable_sizes(diet_vars, "Diet Variables")
+  _log_variable_sizes(diet_vars, "Diet Varaibles")
 
   # Optimize
   total_loss = tf.identity(total_loss, name="total_loss")
diff --git a/tensor2tensor/visualization/TransformerVisualization.ipynb b/tensor2tensor/visualization/TransformerVisualization.ipynb
@@ -15,9 +15,7 @@
   {
    "cell_type": "code",
    "execution_count": 1,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "from __future__ import absolute_import\n",
@@ -36,9 +34,7 @@
   {
    "cell_type": "code",
    "execution_count": 2,
-   "metadata": {
-    "collapsed": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "data": {
@@ -76,9 +72,7 @@
   {
    "cell_type": "code",
    "execution_count": 3,
-   "metadata": {
-    "collapsed": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -111,7 +105,6 @@
    "cell_type": "code",
    "execution_count": 4,
    "metadata": {
-    "collapsed": false,
     "scrolled": true
    },
    "outputs": [
@@ -183,9 +176,7 @@
   {
    "cell_type": "code",
    "execution_count": 6,
-   "metadata": {
-    "collapsed": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -200,15 +191,13 @@
    ],
    "source": [
     "spec = utils.model_builder.model_fn(MODEL, features, tf.estimator.ModeKeys.EVAL, hparams, problem_names=[PROBLEM])\n",
-    "predictions_dict = spec.predictions"
+    "predictions_dict = spec.predictions",
    ]
   },
   {
    "cell_type": "code",
    "execution_count": 7,
-   "metadata": {
-    "collapsed": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -225,7 +214,7 @@
    "source": [
     "with tf.variable_scope(tf.get_variable_scope(), reuse=True):\n",
     "    spec = utils.model_builder.model_fn(MODEL, features, tf.estimator.ModeKeys.PREDICT, hparams, problem_names=[PROBLEM])\n",
-    "    beam_out = spec.predictions['outputs']"
+    "    beam_out = spec.predictions['outputs']",
    ]
   },
   {
@@ -238,9 +227,7 @@
   {
    "cell_type": "code",
    "execution_count": 8,
-   "metadata": {
-    "collapsed": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -320,7 +307,6 @@
    "cell_type": "code",
    "execution_count": 10,
    "metadata": {
-    "collapsed": false,
     "scrolled": false
    },
    "outputs": [
@@ -367,9 +353,7 @@
   {
    "cell_type": "code",
    "execution_count": 12,
-   "metadata": {
-    "collapsed": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -408,9 +392,7 @@
   {
    "cell_type": "code",
    "execution_count": 14,
-   "metadata": {
-    "collapsed": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "data": {
@@ -458,7 +440,6 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": true,
     "scrolled": true
    },
    "outputs": [],
@@ -486,9 +467,9 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython2",
-   "version": "2.7.13"
+   "version": "2.7.12"
   }
  },
  "nbformat": 4,
  "nbformat_minor": 2
-}
+}

Original file line number	Diff line number	Diff line change
`@@ -288,7 +288,7 @@ def nth_model(n):`
`288`	`288`	`diet_vars = [`
`289`	`289`	`v for v in tf.global_variables() if v.dtype == dtypes.float16_ref`
`290`	`290`	`]`
`291`		`- _log_variable_sizes(diet_vars, "Diet Variables")`
	`291`	`+ _log_variable_sizes(diet_vars, "Diet Varaibles")`
`292`	`292`
`293`	`293`	`# Optimize`
`294`	`294`	`total_loss = tf.identity(total_loss, name="total_loss")`