Fix latency benchmark script (#118)

Author: WoosukKwon

benchmark/benchmark_latency.py

@@ -1,71 +1,75 @@
 import argparse
 import time
-from typing import List
 
-from tqdm import tqdm
 import numpy as np
 import torch
+from tqdm import tqdm
 
-from cacheflow.core.server import (
-    add_server_arguments, process_server_arguments,
-    init_local_server_and_frontend_with_arguments)
-from cacheflow.sampling_params import SamplingParams
+from cacheflow import LLM, SamplingParams
 
 
 def main(args: argparse.Namespace):
-    server, frontend = init_local_server_and_frontend_with_arguments(args)
+    print(args)
+
+    # Process all the requests in a single batch if possible.
+    # NOTE(woosuk): If the request cannot be processed in a single batch,
+    # the server will automatically process the request in multiple batches.
+    llm = LLM(
+        model=args.model,
+        tensor_parallel_size=args.tensor_parallel_size,
+        max_num_seqs=args.batch_size,
+        max_num_batched_tokens=args.batch_size * args.input_len,
+    )
 
     sampling_params = SamplingParams(
         n=args.n,
         temperature=0.0 if args.use_beam_search else 1.0,
         top_p=1.0,
         use_beam_search=args.use_beam_search,
-        stop_token_ids=set(),
+        ignore_eos=True,
         max_tokens=args.output_len,
     )
     print(sampling_params)
-    input_token_ids = [0] * args.input_len
+    dummy_prompts = [""] * args.batch_size
+    dummy_prompt_token_ids = [[0] * args.input_len] * args.batch_size
 
-    def profile_step(profile=False):
+    def run_to_completion(profile: bool = False):
         if profile:
             torch.cuda.cudart().cudaProfilerStart()
-        for _ in range(args.batch_size):
-            dummy_prompt = ""
-            frontend._add_query(dummy_prompt, input_token_ids, sampling_params)
-        server.add_sequence_groups(frontend.get_inputs())
         start_time = time.time()
-        while True:
-            server.step()
-            if not server.has_unfinished_requests():
-                break
+
+        llm.generate(dummy_prompts, sampling_params, dummy_prompt_token_ids,
+                     use_tqdm=False)
+
         end_time = time.time()
         latency = end_time - start_time
         if profile:
             torch.cuda.cudart().cudaProfilerStop()
         return latency
 
-    print("Warm up step")
-    profile_step()
+    print("Warming up...")
+    run_to_completion(profile=False)
 
     # Benchmark.
     latencies = []
-    for _ in tqdm(range(3), desc="Profile step"):
-        latencies.append(profile_step())
+    for _ in tqdm(range(args.num_iters), desc="Profiling iterations"):
+        latencies.append(run_to_completion(profile=False))
     print(f'Avg latency: {np.mean(latencies)} seconds')
 
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(
-        description='Benchmark the latency of decoding a single sentence.')
-    parser = add_server_arguments(parser)
+        description='Benchmark the latency of processing a single batch of '
+                    'requests till completion.')
+    parser.add_argument('--model', type=str, default='facebook/opt-125m')
+    parser.add_argument('--tensor-parallel-size', '-tp', type=int, default=1)
     parser.add_argument('--input-len', type=int, default=32)
     parser.add_argument('--output-len', type=int, default=128)
     parser.add_argument('--batch-size', type=int, default=8)
-    parser.add_argument('--n', type=int, default=1)
+    parser.add_argument('--n', type=int, default=1,
+                        help='Number of generated sequences per prompt.')
     parser.add_argument('--use-beam-search', action='store_true')
+    parser.add_argument('--num-iters', type=int, default=3,
+                        help='Number of iterations to run.')
     args = parser.parse_args()
-    args = process_server_arguments(args)
-    args.max_num_batched_tokens = max(
-        args.max_num_batched_tokens, args.batch_size * args.input_len)
-    print(args)
     main(args)

cacheflow/entrypoints/llm.py

@@ -35,18 +35,26 @@ def generate(
         self,
         prompts: List[str],
         sampling_params: Optional[SamplingParams] = None,
+        prompt_token_ids: Optional[List[List[int]]] = None,
         use_tqdm: bool = True,
     ) -> List[RequestOutput]:
         if sampling_params is None:
+            # Use default sampling params.
             sampling_params = SamplingParams()
         # Initialize tqdm.
         if use_tqdm:
             pbar = tqdm(total=len(prompts), desc="Processed prompts")
 
         # Add requests to the server.
-        for prompt in prompts:
+        for i in range(len(prompts)):
+            prompt = prompts[i]
+            if prompt_token_ids is None:
+                token_ids = None
+            else:
+                token_ids = prompt_token_ids[i]
             request_id = str(next(self.request_counter))
-            self.llm_server.add_request(request_id, prompt, sampling_params)
+            self.llm_server.add_request(request_id, prompt, sampling_params,
+                                        token_ids)
 
         # Run the server.
         outputs: List[RequestOutput] = []