Added more measurement info like p50, p90 (#87)

Author: Egor-Krivov

dl_bench/utils.py

@@ -395,34 +395,14 @@ def inference(self, backend: Backend):
         self.net.eval()
         with torch.no_grad():
             start = time.perf_counter()
-            # Duration is inconsistent now
-            with tm.timeit("duration_s"):
-                for i, x in enumerate(test_loader):
-                    backend.sync()
-                    s = get_time()
-                    x = backend.to_device(x)
-                    if backend.dtype != torch.float32:
-                        with torch.autocast(
-                            device_type=backend.device_name,
-                            dtype=backend.dtype,
-                        ):
-                            y = self.net(x)
-                    else:
-                        y = self.net(x)
-
-                    if i < self.warmup_batches:
-                        start = time.perf_counter()
-                        continue
-
-                    backend.sync()
-                    fw_times.append(get_time() - s)
-                    n_items += len(x)
-                    outputs.append(y)
-
-                    # early stopping if we have 10+ batches and were running for 10+ seconds
-                    if (
-                        (time.perf_counter() - start) > self.min_seconds
-                        and n_items >= self.batch_size * self.min_batches
+            for i, x in enumerate(test_loader):
+                backend.sync()
+                s = get_time()
+                x = backend.to_device(x)
+                if backend.dtype != torch.float32:
+                    with torch.autocast(
+                        device_type=backend.device_name,
+                        dtype=backend.dtype,
                     ):
                         y = self.net(x)
                 else: