Update Light-NAS to support latency-aware search (#19050)

* update light_nas_strategy: add latency constraint

test=develop

* update light_nas_strategy: update get_model_latency

test=develop

* update light_nas_strategy: add more check

test=develop

* update light_nas test

test=develop

* update light_nas test

    test=develop

* minor update light_nas test

    test=develop

* minor update light_nas test

test=develop

* update light_nas test

test=develop

* update _constrain_func of light_nas_strategy

test=develop

* update _constrain_func of light_nas_strategy

test=develop

* remove unused code

test=develop

python/paddle/fluid/contrib/slim/nas/light_nas_strategy.py

@@ -40,6 +40,7 @@ class LightNASStrategy(Strategy):
                  controller=None,
                  end_epoch=1000,
                  target_flops=629145600,
+                 target_latency=0,
                  retrain_epoch=1,
                  metric_name='top1_acc',
                  server_ip=None,
@@ -53,6 +54,7 @@ class LightNASStrategy(Strategy):
             controller(searcher.Controller): The searching controller. Default: None.
             end_epoch(int): The 'on_epoch_end' function will be called in end_epoch. Default: 0
             target_flops(int): The constraint of FLOPS.
+            target_latency(float): The constraint of latency.
             retrain_epoch(int): The number of training epochs before evaluating structure generated by controller. Default: 1.
             metric_name(str): The metric used to evaluate the model.
                          It should be one of keys in out_nodes of graph wrapper. Default: 'top1_acc'
@@ -66,6 +68,7 @@ class LightNASStrategy(Strategy):
         self.start_epoch = 0
         self.end_epoch = end_epoch
         self._max_flops = target_flops
+        self._max_latency = target_latency
         self._metric_name = metric_name
         self._controller = controller
         self._retrain_epoch = 0
@@ -86,8 +89,6 @@ class LightNASStrategy(Strategy):
 
     def on_compression_begin(self, context):
         self._current_tokens = context.search_space.init_tokens()
-        constrain_func = functools.partial(
-            self._constrain_func, context=context)
         self._controller.reset(context.search_space.range_table(),
                                self._current_tokens, None)
 
@@ -127,15 +128,6 @@ class LightNASStrategy(Strategy):
                 d[key] = self.__dict__[key]
         return d
 
-    def _constrain_func(self, tokens, context=None):
-        """Check whether the tokens meet constraint."""
-        _, _, test_prog, _, _, _, _ = context.search_space.create_net(tokens)
-        flops = GraphWrapper(test_prog).flops()
-        if flops <= self._max_flops:
-            return True
-        else:
-            return False
-
     def on_epoch_begin(self, context):
         if context.epoch_id >= self.start_epoch and context.epoch_id <= self.end_epoch and (
                 self._retrain_epoch == 0 or
@@ -144,13 +136,20 @@ class LightNASStrategy(Strategy):
             for _ in range(self._max_try_times):
                 startup_p, train_p, test_p, _, _, train_reader, test_reader = context.search_space.create_net(
                     self._current_tokens)
-                _logger.info("try [{}]".format(self._current_tokens))
                 context.eval_graph.program = test_p
                 flops = context.eval_graph.flops()
-                if flops <= self._max_flops:
-                    break
+                if self._max_latency > 0:
+                    latency = context.search_space.get_model_latency(test_p)
+                    _logger.info("try [{}] with latency {} flops {}".format(
+                        self._current_tokens, latency, flops))
                 else:
+                    _logger.info("try [{}] with flops {}".format(
+                        self._current_tokens, flops))
+                if flops > self._max_flops or (self._max_latency > 0 and
+                                               latency > self._max_latency):
                     self._current_tokens = self._search_agent.next_tokens()
+                else:
+                    break
 
             context.train_reader = train_reader
             context.eval_reader = test_reader
@@ -173,6 +172,16 @@ class LightNASStrategy(Strategy):
             flops = context.eval_graph.flops()
             if flops > self._max_flops:
                 self._current_reward = 0.0
+            if self._max_latency > 0:
+                test_p = context.search_space.create_net(self._current_tokens)[
+                    2]
+                latency = context.search_space.get_model_latency(test_p)
+                if latency > self._max_latency:
+                    self._current_reward = 0.0
+                _logger.info("reward: {}; latency: {}; flops: {}; tokens: {}".
+                             format(self._current_reward, latency, flops,
+                                    self._current_tokens))
+            else:
                 _logger.info("reward: {}; flops: {}; tokens: {}".format(
                     self._current_reward, flops, self._current_tokens))
             self._current_tokens = self._search_agent.update(

python/paddle/fluid/contrib/slim/nas/search_space.py

@@ -41,3 +41,12 @@ class SearchSpace(object):
             (tuple): startup_program, train_program, evaluation_program, train_metrics, test_metrics
         """
         raise NotImplementedError('Abstract method.')
+
+    def get_model_latency(self, program):
+        """Get model latency according to program.
+        Args:
+            program(Program): The program to get latency.
+        Return:
+            (float): model latency.
+        """
+        raise NotImplementedError('Abstract method.')

python/paddle/fluid/contrib/slim/tests/light_nas/compress.yaml

@@ -10,6 +10,7 @@ strategies:
         class: 'LightNASStrategy'
         controller: 'sa_controller'
         target_flops: 629145600
+        target_latency: 1
         end_epoch: 2
         retrain_epoch: 1
         metric_name: 'acc_top1'

python/paddle/fluid/contrib/slim/tests/light_nas/light_nas_space.py

@@ -17,6 +17,7 @@ from light_nasnet import LightNASNet
 import paddle.fluid as fluid
 import paddle
 import json
+import random
 
 total_images = 1281167
 lr = 0.1
@@ -85,6 +86,16 @@ class LightNASSpace(SearchSpace):
             2, 4, 3, 3, 2, 2, 2
         ]
 
+    def get_model_latency(self, program):
+        """Get model latency according to program.
+        Returns a random number since it's only for testing.
+        Args:
+            program(Program): The program to get latency.
+        Return:
+            (float): model latency.
+        """
+        return random.randint(1, 2)
+
     def create_net(self, tokens=None):
         """Create a network for training by tokens.
         """

python/paddle/fluid/contrib/slim/tests/test_light_nas.py

@@ -11,24 +11,96 @@
 # 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.
-
-import paddle
+"""
+Test LightNAS.
+"""
+import sys
 import unittest
 import paddle.fluid as fluid
-from mobilenet import MobileNet
 from paddle.fluid.contrib.slim.core import Compressor
-from paddle.fluid.contrib.slim.graph import GraphWrapper
-import sys
 sys.path.append("./light_nas")
 from light_nas_space import LightNASSpace
 
 
 class TestLightNAS(unittest.TestCase):
+    """
+    Test LightNAS.
+    """
+
     def test_compression(self):
+        """
+        Test LightNAS.
+        """
+        # Update compress.yaml
+        lines = list()
+        fid = open('./light_nas/compress.yaml')
+        for line in fid:
+            if 'target_latency' in line:
+                lines.append('        target_latency: 0\n')
+            else:
+                lines.append(line)
+        fid.close()
+        fid = open('./light_nas/compress.yaml', 'w')
+        for line in lines:
+            fid.write(line)
+        fid.close()
+
+        # Begin test
+        if not fluid.core.is_compiled_with_cuda():
+            return
+
+        space = LightNASSpace()
+
+        startup_prog, train_prog, test_prog, train_metrics, test_metrics, train_reader, test_reader = space.create_net(
+        )
+        train_cost, train_acc1, train_acc5, global_lr = train_metrics
+        test_cost, test_acc1, test_acc5 = test_metrics
+
+        place = fluid.CUDAPlace(0)
+        exe = fluid.Executor(place)
+        exe.run(startup_prog)
+
+        val_fetch_list = [('acc_top1', test_acc1.name),
+                          ('acc_top5', test_acc5.name)]
+        train_fetch_list = [('loss', train_cost.name)]
+
+        com_pass = Compressor(
+            place,
+            fluid.global_scope(),
+            train_prog,
+            train_reader=train_reader,
+            train_feed_list=None,
+            train_fetch_list=train_fetch_list,
+            eval_program=test_prog,
+            eval_reader=test_reader,
+            eval_feed_list=None,
+            eval_fetch_list=val_fetch_list,
+            train_optimizer=None,
+            search_space=space)
+        com_pass.config('./light_nas/compress.yaml')
+        eval_graph = com_pass.run()
+
+    def test_compression_with_target_latency(self):
+        """
+        Test LightNAS with target_latency.
+        """
+        # Update compress.yaml
+        lines = list()
+        fid = open('./light_nas/compress.yaml')
+        for line in fid:
+            if 'target_latency' in line:
+                lines.append('        target_latency: 1\n')
+            else:
+                lines.append(line)
+        fid.close()
+        fid = open('./light_nas/compress.yaml', 'w')
+        for line in lines:
+            fid.write(line)
+        fid.close()
+
+        # Begin test
         if not fluid.core.is_compiled_with_cuda():
             return
-        class_dim = 10
-        image_shape = [1, 28, 28]
 
         space = LightNASSpace()
 
@@ -41,8 +113,8 @@ class TestLightNAS(unittest.TestCase):
         exe = fluid.Executor(place)
         exe.run(startup_prog)
 
-        val_fetch_list = [('acc_top1', test_acc1.name), ('acc_top5',
-                                                         test_acc5.name)]
+        val_fetch_list = [('acc_top1', test_acc1.name),
+                          ('acc_top5', test_acc5.name)]
         train_fetch_list = [('loss', train_cost.name)]
 
         com_pass = Compressor(