Fix bug in TFLite GPU parser (#92)

nn_meter/builder/backends/tflite/gpu.py

@@ -18,27 +18,8 @@ class TFLiteGPULatencyParser(BaseParser):
         self.after_fused_graph = ''
 
     def parse(self, content):
-        result = self._parse_time(content)
-        kernel_operation_map = self._parse_kernel_name(content)
-        work_size = self._parse_work_size(content)
-        self.realtime, self.block_name = self._parse_block(content)
-        self.kernel_sum = sum(value[0] for key, value in result.items())
-        self.kernels = [{}] * len(result)
-        self.before_fused_graph, self.after_fused_graph = self._parse_graph(content)
         self.comp_avg, self.comp_std = self._parse_comp_time(content)
-        self.nodes = self._parse_node_cpu_time(content)
-        self.errors = self._parse_error(content)
-        for key, value in result.items():
-            self.kernels[key] = {
-                'avg': value[0],
-                'std': value[1],
-                'work_size': work_size[key],
-                'name': kernel_operation_map[key],
-            }
-
-        self.comp_kernel_latency = sum((Latency(kernel['avg'], kernel['std']) for kernel in self.kernels if kernel['name'] != 'to/from tensor'), Latency())
-
-        self.raw_content = content
+        self.comp_kernel_latency = Latency(self.comp_avg, self.comp_std)
 
         return self
 
@@ -134,7 +115,7 @@ class TFLiteGPULatencyParser(BaseParser):
         return before_fused_graph, after_fused_graph
 
     def _parse_comp_time(self, content):
-        comp_time_regex = r'comp_avg_ms=([\d.e-]+) comp_std_ms=([\d.e-]+)'
+        comp_time_regex = r'comp_avg_ms=([\d.\+e-]+) comp_std_ms=([\d.\+e-]+)'
         comp_avg, comp_std = 0, 0
 
         for line in content.splitlines():

nn_meter/builder/backends/tflite/tflite_profiler.py

@@ -50,7 +50,6 @@ class TFLiteProfiler(BaseProfiler):
             kernel_cmd = f'--kernel_path={self._dst_kernel_path}' if self._dst_kernel_path else ''
             close_xnnpack_cmd = f'--use_xnnpack=false' if close_xnnpack else ''
             res = device.shell(f' {taskset_cmd} {self._benchmark_model_path} {kernel_cmd} {close_xnnpack_cmd}' \
-                               f' {close_xnnpack_cmd}' \
                                f' --num_threads={self._num_threads}' \
                                f' --num_runs={self._num_runs}' \
                                f' --warmup_runs={self._warm_ups}' \

tests/integration_test/data/reference_result_onnx_based_torch.txt

@@ -3,7 +3,7 @@ resnet18, torch, cortexA76cpu_tflite21, 1.0, 216.1971
 alexnet, torch, cortexA76cpu_tflite21, 1.0, 96.5713
 vgg16, torch, cortexA76cpu_tflite21, 1.0, 1668.5085
 squeezenet, torch, cortexA76cpu_tflite21, 1.0, 98.3766
-densenet161, torch, cortexA76cpu_tflite21, 1.0, 917.2059
+densenet161, torch, cortexA76cpu_tflite21, 1.0, 917.0045
 inception_v3, torch, cortexA76cpu_tflite21, 1.0, 586.2565
 googlenet, torch, cortexA76cpu_tflite21, 1.0, 167.5816
 shufflenet_v2, torch, cortexA76cpu_tflite21, 1.0, 21.3181
@@ -15,7 +15,7 @@ resnet18, torch, adreno640gpu_tflite21, 1.0, 39.3235
 alexnet, torch, adreno640gpu_tflite21, 1.0, 13.1267
 vgg16, torch, adreno640gpu_tflite21, 1.0, 219.2648
 squeezenet, torch, adreno640gpu_tflite21, 1.0, 18.6742
-densenet161, torch, adreno640gpu_tflite21, 1.0, 186.5604
+densenet161, torch, adreno640gpu_tflite21, 1.0, 186.4237
 inception_v3, torch, adreno640gpu_tflite21, 1.0, 127.9842
 googlenet, torch, adreno640gpu_tflite21, 1.0, 32.7581
 shufflenet_v2, torch, adreno640gpu_tflite21, 1.0, 5.4239
@@ -27,7 +27,7 @@ resnet18, torch, adreno630gpu_tflite21, 1.0, 49.4287
 alexnet, torch, adreno630gpu_tflite21, 1.0, 16.8673
 vgg16, torch, adreno630gpu_tflite21, 1.0, 286.5998
 squeezenet, torch, adreno630gpu_tflite21, 1.0, 21.095
-densenet161, torch, adreno630gpu_tflite21, 1.0, 193.0796
+densenet161, torch, adreno630gpu_tflite21, 1.0, 192.9225
 inception_v3, torch, adreno630gpu_tflite21, 1.0, 161.3187
 googlenet, torch, adreno630gpu_tflite21, 1.0, 37.9612
 shufflenet_v2, torch, adreno630gpu_tflite21, 1.0, 5.5792