[V1] fix torch profiling for V1 offline scenarios (vllm-project#18445)

Signed-off-by: Divakar Verma <[email protected]>

Author: divakar-amd

benchmarks/benchmark_latency.py

@@ -6,13 +6,12 @@
 import json
 import os
 import time
-from pathlib import Path
 from typing import Any, Optional
 
 import numpy as np
-import torch
 from tqdm import tqdm
 
+import vllm.envs as envs
 from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
 from vllm import LLM, SamplingParams
 from vllm.engine.arg_utils import EngineArgs
@@ -80,17 +79,9 @@ def llm_generate():
 
     def run_to_completion(profile_dir: Optional[str] = None):
         if profile_dir:
-            with torch.profiler.profile(
-                activities=[
-                    torch.profiler.ProfilerActivity.CPU,
-                    torch.profiler.ProfilerActivity.CUDA,
-                ],
-                on_trace_ready=torch.profiler.tensorboard_trace_handler(
-                    str(profile_dir)
-                ),
-            ) as p:
-                llm_generate()
-            print(p.key_averages().table(sort_by="self_cuda_time_total"))
+            llm.start_profile()
+            llm_generate()
+            llm.stop_profile()
         else:
             start_time = time.perf_counter()
             llm_generate()
@@ -103,11 +94,7 @@ def run_to_completion(profile_dir: Optional[str] = None):
         run_to_completion(profile_dir=None)
 
     if args.profile:
-        profile_dir = args.profile_result_dir
-        if not profile_dir:
-            profile_dir = (
-                Path(".") / "vllm_benchmark_result" / f"latency_result_{time.time()}"
-            )
+        profile_dir = envs.VLLM_TORCH_PROFILER_DIR
         print(f"Profiling (results will be saved to '{profile_dir}')...")
         run_to_completion(profile_dir=profile_dir)
         return
@@ -164,15 +151,6 @@ def run_to_completion(profile_dir: Optional[str] = None):
         action="store_true",
         help="profile the generation process of a single batch",
     )
-    parser.add_argument(
-        "--profile-result-dir",
-        type=str,
-        default=None,
-        help=(
-            "path to save the pytorch profiler output. Can be visualized "
-            "with ui.perfetto.dev or Tensorboard."
-        ),
-    )
     parser.add_argument(
         "--output-json",
         type=str,
@@ -193,4 +171,9 @@ def run_to_completion(profile_dir: Optional[str] = None):
     # numbers. We need to disable prefix caching by default.
     parser.set_defaults(enable_prefix_caching=False)
     args = parser.parse_args()
+    if args.profile and not envs.VLLM_TORCH_PROFILER_DIR:
+        raise OSError(
+            "The environment variable 'VLLM_TORCH_PROFILER_DIR' is not set. "
+            "Please set it to a valid path to use torch profiler."
+        )
     main(args)

vllm/benchmarks/latency.py

@@ -6,13 +6,12 @@
 import json
 import os
 import time
-from pathlib import Path
 from typing import Any, Optional
 
 import numpy as np
-import torch
 from tqdm import tqdm
 
+import vllm.envs as envs
 from vllm import LLM, SamplingParams
 from vllm.benchmarks.utils import (convert_to_pytorch_benchmark_format,
                                    write_to_json)
@@ -59,13 +58,6 @@ def add_cli_args(parser: argparse.ArgumentParser):
         action="store_true",
         help="profile the generation process of a single batch",
     )
-    parser.add_argument(
-        "--profile-result-dir",
-        type=str,
-        default=None,
-        help=("path to save the pytorch profiler output. Can be visualized "
-              "with ui.perfetto.dev or Tensorboard."),
-    )
     parser.add_argument(
         "--output-json",
         type=str,
@@ -87,7 +79,10 @@ def add_cli_args(parser: argparse.ArgumentParser):
 
 def main(args: argparse.Namespace):
     print(args)
-
+    if args.profile and not envs.VLLM_TORCH_PROFILER_DIR:
+        raise OSError(
+            "The environment variable 'VLLM_TORCH_PROFILER_DIR' is not set. "
+            "Please set it to a valid path to use torch profiler.")
     engine_args = EngineArgs.from_cli_args(args)
 
     # NOTE(woosuk): If the request cannot be processed in a single batch,
@@ -131,16 +126,9 @@ def llm_generate():
 
     def run_to_completion(profile_dir: Optional[str] = None):
         if profile_dir:
-            with torch.profiler.profile(
-                    activities=[
-                        torch.profiler.ProfilerActivity.CPU,
-                        torch.profiler.ProfilerActivity.CUDA,
-                    ],
-                    on_trace_ready=torch.profiler.tensorboard_trace_handler(
-                        str(profile_dir)),
-            ) as p:
-                llm_generate()
-            print(p.key_averages().table(sort_by="self_cuda_time_total"))
+            llm.start_profile()
+            llm_generate()
+            llm.stop_profile()
         else:
             start_time = time.perf_counter()
             llm_generate()
@@ -153,10 +141,7 @@ def run_to_completion(profile_dir: Optional[str] = None):
         run_to_completion(profile_dir=None)
 
     if args.profile:
-        profile_dir = args.profile_result_dir
-        if not profile_dir:
-            profile_dir = (Path(".") / "vllm_benchmark_result" /
-                           f"latency_result_{time.time()}")
+        profile_dir = envs.VLLM_TORCH_PROFILER_DIR
         print(f"Profiling (results will be saved to '{profile_dir}')...")
         run_to_completion(profile_dir=profile_dir)
         return

vllm/v1/worker/gpu_worker.py

@@ -292,6 +292,8 @@ def profile(self, is_start: bool = True):
             self.profiler.start()
         else:
             self.profiler.stop()
+            print(self.profiler.key_averages().table(
+                sort_by="self_cuda_time_total"))
 
     def execute_dummy_batch(self) -> None:
         self.model_runner._dummy_run(1)

vllm/worker/worker.py

@@ -128,6 +128,8 @@ def stop_profile(self):
         if self.profiler is None:
             raise RuntimeError("Profiler is not enabled.")
         self.profiler.stop()
+        print(
+            self.profiler.key_averages().table(sort_by="self_cuda_time_total"))
 
     def sleep(self, level: int = 1) -> None:
         free_bytes_before_sleep = torch.cuda.mem_get_info()[0]