Merge 5a65b93 into 245741e

oss_scripts/oss_tests.sh

@@ -62,6 +62,7 @@ pytest --disable-warnings \
   --ignore=trax/supervised/trainer_lib_test.py \
   --ignore=trax/supervised/training_test.py \
   --ignore=trax/supervised/decoding_test.py \
+  --ignore=trax/supervised/decoding_timing_test.py \
   trax/supervised
 set_status
 

trax/supervised/decoding_test.py

@@ -18,7 +18,6 @@
 
 import functools
 import os
-import time
 
 from jax import test_util  # pylint: disable=unused-import
 from jax.config import config
@@ -202,61 +201,6 @@ def test_autoregressive_sample_reformer2_lsh(self):
     self.assertEqual(s.shape[0], 1)
     self.assertEqual(s.shape[1], 10)
 
-  def test_autoregressive_sample_reformer2_timing(self):
-    max_len = 16
-
-    def _self_attention_fn():
-      return functools.partial(
-          layers.SelfAttention,
-          predict_drop_len=2 * max_len,
-          predict_mem_len=2 * max_len)
-
-    def _causal_attention_fn():
-      return functools.partial(
-          layers.CausalAttention,
-          max_inference_length=2 * max_len)
-
-    pred_model = models.Reformer2(
-        mode='predict',
-        d_model=4*1024,
-        d_ff=32*1024,
-        dropout=0.05,
-        max_len=max_len,
-        n_heads=16,
-        n_encoder_layers=3,
-        n_decoder_layers=3,
-        encoder_attention_type=_self_attention_fn(),
-        encoder_decoder_attention_type=_causal_attention_fn(),
-        input_vocab_size=32,
-        ff_sparsity=128,
-        axial_pos_shape=None,
-    )
-
-    shape11 = shapes.ShapeDtype((1, 1), dtype=np.int32)
-    shape1l = shapes.ShapeDtype((1, max_len), dtype=np.int32)
-    pred_model.init(input_signature=(shape1l, shape11))
-    inputs = np.arange(16, dtype=np.int32).reshape(1, 16)
-
-    # This is decoding.autoregressive_sample but simplified and with timing.
-    result, counter, start_time, total_time = [], 0, time.time(), 0.0
-    for sample in decoding.autoregressive_sample_stream(
-        pred_model, inputs, temperature=0.0):  # accelerate=False):
-      elapsed_time = time.time() - start_time
-      start_time = time.time()
-      if counter > 3:
-        total_time += elapsed_time
-      result.append(sample[:, None])
-      counter += 1
-      if counter >= 14:
-        break
-
-    print('\n\n\nTotal time (10 tokens): %.4fs\n\n\n' % total_time)
-    self.assertLess(total_time, 10.0)  # If it's > 10s, it's some bug.
-    # Check resulting shapes.
-    s = np.concatenate(result, axis=1)
-    self.assertEqual(s.shape[0], 1)
-    self.assertEqual(s.shape[1], 14)
-
   def test_autoregressive_sample_reformer2_copy_self_attn_quality(self):
     max_len = 32
 

trax/supervised/decoding_timing_test.py

@@ -0,0 +1,93 @@
+# coding=utf-8
+# Copyright 2020 The Trax Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# Lint as: python3
+"""Timing tests for decoding."""
+
+import functools
+import time
+
+from jax import test_util  # pylint: disable=unused-import
+from jax.config import config
+import numpy as np
+from tensorflow.compat.v2 import test
+
+from trax import layers
+from trax import models
+from trax import shapes
+from trax.supervised import decoding
+
+
+class DecodingTimingTest(test.TestCase):
+
+  def test_autoregressive_sample_reformer2_timing(self):
+    max_len = 16
+
+    def _self_attention_fn():
+      return functools.partial(
+          layers.SelfAttention,
+          predict_drop_len=2 * max_len,
+          predict_mem_len=2 * max_len)
+
+    def _causal_attention_fn():
+      return functools.partial(
+          layers.CausalAttention,
+          max_inference_length=2 * max_len)
+
+    pred_model = models.Reformer2(
+        mode='predict',
+        d_model=4*1024,
+        d_ff=32*1024,
+        dropout=0.05,
+        max_len=max_len,
+        n_heads=16,
+        n_encoder_layers=3,
+        n_decoder_layers=3,
+        encoder_attention_type=_self_attention_fn(),
+        encoder_decoder_attention_type=_causal_attention_fn(),
+        input_vocab_size=32,
+        ff_sparsity=128,
+        axial_pos_shape=None,
+    )
+
+    shape11 = shapes.ShapeDtype((1, 1), dtype=np.int32)
+    shape1l = shapes.ShapeDtype((1, max_len), dtype=np.int32)
+    pred_model.init(input_signature=(shape1l, shape11))
+    inputs = np.arange(16, dtype=np.int32).reshape(1, 16)
+
+    # This is decoding.autoregressive_sample but simplified and with timing.
+    result, counter, start_time, total_time = [], 0, time.time(), 0.0
+    for sample in decoding.autoregressive_sample_stream(
+        pred_model, inputs, temperature=0.0):  # accelerate=False):
+      elapsed_time = time.time() - start_time
+      start_time = time.time()
+      if counter > 3:
+        total_time += elapsed_time
+      result.append(sample[:, None])
+      counter += 1
+      if counter >= 14:
+        break
+
+    print('\n\n\nTotal time (10 tokens): %.4fs\n\n\n' % total_time)
+    self.assertLess(total_time, 20.0)  # If it's > 20s, it's some bug.
+    # Check resulting shapes.
+    s = np.concatenate(result, axis=1)
+    self.assertEqual(s.shape[0], 1)
+    self.assertEqual(s.shape[1], 14)
+
+
+if __name__ == '__main__':
+  config.config_with_absl()
+  test.main()