Method of adding hardware (PaddlePaddle#1041)

* fix opt's cmd for sparse model

* add hardware

* Remove redundant functions and adjust tests' file

* Remove redundant functions and adjust tests' file

Author: zzjjay

paddleslim/analysis/__init__.py

@@ -16,18 +16,10 @@
 from .latency import LatencyEvaluator, TableLatencyEvaluator
 from .latency_predictor import LatencyPredictor, TableLatencyPredictor
 from .parse_ops import get_key_from_op
-from ._utils import save_cls_model, save_det_model, save_seg_model
+from ._utils import save_cls_model, save_det_model
 
 __all__ = [
-    'flops',
-    'dygraph_flops',
-    'model_size',
-    'LatencyEvaluator',
-    'TableLatencyEvaluator',
-    "LatencyPredictor",
-    "TableLatencyPredictor",
-    "get_key_from_op",
-    "save_cls_model",
-    "save_det_model",
-    "save_seg_model",
+    'flops', 'dygraph_flops', 'model_size', 'LatencyEvaluator',
+    'TableLatencyEvaluator', "LatencyPredictor", "TableLatencyPredictor",
+    "get_key_from_op", "save_cls_model", "save_det_model"
 ]

paddleslim/analysis/_utils.py

@@ -20,8 +20,8 @@
 import subprocess
 import time
 __all__ = [
-    "save_cls_model", "save_det_model", "save_seg_model", "nearest_interpolate",
-    "opt_model", "load_predictor"
+    "save_cls_model", "save_det_model", "nearest_interpolate", "opt_model",
+    "load_predictor"
 ]
 
 
@@ -30,8 +30,7 @@ def opt_model(opt="paddle_lite_opt",
               param_file='',
               optimize_out_type='protobuf',
               valid_targets='arm',
-              enable_fp16=False,
-              sparse_ratio=0):
+              enable_fp16=False):
     assert os.path.exists(model_file) and os.path.exists(
         param_file), f'{model_file} or {param_file} does not exist.'
     save_dir = f'./opt_models_tmp/{os.getpid()}_{time.time()}'
@@ -40,15 +39,13 @@ def opt_model(opt="paddle_lite_opt",
 
     assert optimize_out_type in ['protobuf', 'naive_buffer']
     if optimize_out_type == 'protobuf':
-        model_out = os.path.join(save_dir, 'pbmodel')
+        model_out = save_dir
     else:
         model_out = os.path.join(save_dir, 'model')
 
     enable_fp16 = str(enable_fp16).lower()
-    sparse_model = True if sparse_ratio > 0 else False
-    sparse_threshold = max(sparse_ratio - 0.1, 0.1)
 
-    cmd = f'{opt} --model_file={model_file} --param_file={param_file}  --optimize_out_type={optimize_out_type} --optimize_out={model_out} --valid_targets={valid_targets} --enable_fp16={enable_fp16} --sparse_model={sparse_model} --sparse_threshold={sparse_threshold}'
+    cmd = f'{opt} --model_file={model_file} --param_file={param_file}  --optimize_out_type={optimize_out_type} --optimize_out={model_out} --valid_targets={valid_targets} --enable_fp16={enable_fp16} --sparse_model=true --sparse_threshold=0.4'
     print(f'commands:{cmd}')
     m = subprocess.Popen(
         cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True)
@@ -169,48 +166,6 @@ def save_det_model(model,
     return model_file, param_file
 
 
-def save_seg_model(model, input_shape, save_dir, data_type):
-    if data_type == 'fp32':
-        paddle.jit.save(
-            model,
-            path=os.path.join(save_dir, 'fp32model'),
-            input_spec=[
-                paddle.static.InputSpec(
-                    shape=input_shape, dtype='float32', name='x'),
-            ])
-        model_file = os.path.join(save_dir, 'fp32model.pdmodel')
-        param_file = os.path.join(save_dir, 'fp32model.pdiparams')
-
-    else:
-        save_dir = os.path.join(save_dir, 'int8model')
-        quant_config = {
-            'weight_preprocess_type': None,
-            'activation_preprocess_type': None,
-            'weight_quantize_type': 'channel_wise_abs_max',
-            'activation_quantize_type': 'moving_average_abs_max',
-            'weight_bits': 8,
-            'activation_bits': 8,
-            'dtype': 'int8',
-            'window_size': 10000,
-            'moving_rate': 0.9,
-            'quantizable_layer_type': ['Conv2D', 'Linear'],
-        }
-        quantizer = paddleslim.QAT(config=quant_config)
-        quantizer.quantize(model)
-        quantizer.save_quantized_model(
-            model,
-            save_dir,
-            input_spec=[
-                paddle.static.InputSpec(
-                    shape=input_shape, dtype='float32')
-            ])
-
-        model_file = f'{save_dir}.pdmodel'
-        param_file = f'{save_dir}.pdiparams'
-
-    return model_file, param_file
-
-
 def nearest_interpolate(features, data):
     def distance(x, y):
         x = np.array(x)

paddleslim/analysis/latency_predictor.py

@@ -39,7 +39,7 @@ class LatencyPredictor(object):
     """Base class of latency predictor.
     """
 
-    def predict_latency(self, model):
+    def predict(self, model):
         """Get latency of model. It is an abstract method.
 
         Args:
@@ -64,6 +64,7 @@ class TableLatencyPredictor(LatencyPredictor):
     Args:
         table_file(str): The path of file that records the device latency of operators.
     """
+    hardware_list = ['SD625', 'SD710']
 
     def __init__(self, table_file='SD710'):
         self.table_file = table_file
@@ -72,11 +73,14 @@ def __init__(self, table_file='SD710'):
         self.threads = None
         self.predictor_state = False
         self.predictor = {}
-        self.hardware_list = ['SD625', 'SD710']
         self._initial_table()
 
+    @classmethod
+    def add_hardware(cls, hardware):
+        cls.hardware_list.append(hardware)
+
     def _initial_table(self):
-        if self.table_file in self.hardware_list:
+        if self.table_file in TableLatencyPredictor.hardware_list:
             self.hardware = self.table_file
             self.threads = 4
             self.table_file = f'{self.hardware}_threads_4_power_mode_0.pkl'
@@ -88,7 +92,7 @@ def _initial_table(self):
 
         assert os.path.exists(
             self.table_file
-        ), f'{self.table_file} does not exist. If you want to use our table files, please set \'table_file\' in {self.hardware_list}'
+        ), f'{self.table_file} does not exist. If you want to use our table files, please set \'table_file\' in {TableLatencyPredictor.hardware_list}'
         with open(self.table_file, 'rb') as f:
             self.table_dict = pickle.load(f)
 
@@ -123,6 +127,8 @@ def _preload_predictor(self, data_type='fp32'):
         ]
         op_dir = self.table_file.split('.')[0] + '_batchsize_1'
         for op_type in op_types:
+            if data_type == 'fp32' and op_type == 'calib':
+                continue
             model = load_predictor(op_type, op_dir, data_type)
             key = op_type
             if 'conv2d' in op_type:
@@ -141,8 +147,6 @@ def predict(self,
             model_file(str), param_file(str): The inference model(*.pdmodel, *.pdiparams).
             data_type(str): Data type, fp32, fp16 or int8.
             threads(int): Threads num.
-            sparse_ratio(float): The ratio of unstructured pruning.
-            prune_ratio(float): The ration of structured pruning.
             input_shape(list): Generally, the input shape is confirmed when saving the inference model and the parameter is only effective for input shape that has variable length.
         Returns:
             latency(float): The latency of the model.

tests/test_latency_predictor.py

@@ -19,7 +19,7 @@
 from paddleslim.analysis import LatencyPredictor, TableLatencyPredictor
 from paddle.vision.models import mobilenet_v1, mobilenet_v2
 from paddle.nn import Conv2D, BatchNorm2D, ReLU, LayerNorm
-from paddleslim.analysis._utils import opt_model, save_cls_model, save_seg_model, save_det_model
+from paddleslim.analysis._utils import opt_model, save_cls_model, save_det_model
 
 
 def channel_shuffle(x, groups):
@@ -276,7 +276,7 @@ def test_case5(self):
         paddle.disable_static()
         model = mobilenet_v1()
         predictor = TableLatencyPredictor(table_file='SD710')
-        model_file, param_file = save_seg_model(
+        model_file, param_file = save_cls_model(
             model,
             input_shape=[1, 3, 224, 224],
             save_dir="./inference_model",
@@ -367,30 +367,6 @@ def test_case9(self):
 
 class TestCase10(unittest.TestCase):
     def test_case10(self):
-        paddle.disable_static()
-        model = ModelCase1()
-        predictor = LatencyPredictor()
-        model_file, param_file = save_seg_model(
-            model,
-            input_shape=[1, 116, 28, 28],
-            save_dir="./inference_model",
-            data_type='int8')
-        pbmodel_file = opt_model(
-            model_file=model_file,
-            param_file=param_file,
-            optimize_out_type='protobuf')
-
-        paddle.enable_static()
-        with open(pbmodel_file, "rb") as f:
-            fluid_program = paddle.fluid.framework.Program.parse_from_string(
-                f.read())
-            graph = paddleslim.core.GraphWrapper(fluid_program)
-            graph_keys = predictor._get_key_info_from_graph(graph=graph)
-            assert len(graph_keys) > 0
-
-
-class TestCase11(unittest.TestCase):
-    def test_case11(self):
         paddle.disable_static()
         model = mobilenet_v2()
         model2 = ModelCase6()