Make example name less confusing

examples/cuda-benchmark/cpu_bench.yml

@@ -4,11 +4,11 @@ nodes:
       next: node_2/next
     outputs:
       - latency
-    path: cuda_sender.py
+    path: sender.py
     env:
       DEVICE: cpu
   - id: node_2
-    path: cuda_receiver.py
+    path: receiver.py
     inputs:
       next: node_1/latency
     outputs:

examples/cuda-benchmark/cuda_bench.yml

@@ -4,9 +4,9 @@ nodes:
       next: node_2/next
     outputs:
       - latency
-    path: cuda_sender.py
+    path: sender.py
   - id: node_2
-    path: cuda_receiver.py
+    path: receiver.py
     inputs:
       next: node_1/latency
     outputs:

examples/cuda-benchmark/cuda_receiver.py → examples/cuda-benchmark/receiver.py

@@ -46,12 +46,10 @@ while True:
             ipc_handle = ipc_buffer_to_ipc_handle(event["value"])
             cudabuffer = ctx.open_ipc_buffer(ipc_handle)
             torch_tensor = cudabuffer_to_torch(cudabuffer, event["metadata"])  # on cuda
-            print(torch_tensor[0])
     else:
         break
     t_received = time.perf_counter_ns()
     length = len(torch_tensor) * 8
-    pbar.update(1)
 
     if length != current_size:
         if n > 0:
@@ -64,6 +62,7 @@ while True:
         start = time.perf_counter_ns()
         latencies = []
 
+    pbar.update(1)
     latencies.append((t_received - t_send) / 1000)
     node.send_output("next", pa.array([]))
 

examples/cuda-benchmark/cuda_sender.py → examples/cuda-benchmark/sender.py

@@ -11,7 +11,7 @@ import torch
 
 torch.tensor([], device="cuda")
 
-SIZES = [512, 10 * 512, 100 * 512, 1000 * 512, 10000 * 512]
+SIZES = [10000 * 512]
 
 DEVICE = os.getenv("DEVICE", "cuda")
 
@@ -43,4 +43,3 @@ for size in SIZES:
 
         # Wait before sending next output
         node.next()
-        time.sleep(0.1)