fix: add seed bench_serving to cache key, remove redundant function definition. (#12680)

This commit is contained in:
yinghui
2025-11-04 23:39:11 -08:00
committed by GitHub
parent 3694266051
commit b88fab3111
2 changed files with 1 additions and 22 deletions
-21
View File
@@ -2,7 +2,6 @@ import json
import os import os
import pickle import pickle
import random import random
from pathlib import Path
from typing import List, Optional, Tuple, Union from typing import List, Optional, Tuple, Union
import numpy as np import numpy as np
@@ -426,26 +425,6 @@ def sample_random_requests(
return input_requests return input_requests
def gen_prompt(tokenizer, token_num):
"""Generate a random prompt of specified token length using tokenizer vocabulary."""
all_available_tokens = list(tokenizer.get_vocab().values())
selected_tokens = random.choices(all_available_tokens, k=token_num)
return tokenizer.decode(selected_tokens)
def get_gen_prefix_cache_path(args, tokenizer):
"""Create cache directory under ~/.cache/sglang/benchmark"""
cache_dir = Path.home() / ".cache" / "sglang" / "benchmark"
# Create a unique cache filename based on the generation parameters
cache_key = (
f"gsp_prefix_{args.gsp_num_groups}_{args.gsp_prompts_per_group}_"
f"{args.gsp_system_prompt_len}_{args.gsp_question_len}_{args.gsp_output_len}_"
f"{tokenizer.__class__.__name__}.pkl"
)
return cache_dir / cache_key
def sample_generated_shared_prefix_requests( def sample_generated_shared_prefix_requests(
num_groups: int, num_groups: int,
prompts_per_group: int, prompts_per_group: int,
+1 -1
View File
@@ -1507,7 +1507,7 @@ def get_gen_prefix_cache_path(args, tokenizer):
# Create a unique cache filename based on the generation parameters # Create a unique cache filename based on the generation parameters
cache_key = ( cache_key = (
f"gen_shared_prefix_{args.gsp_num_groups}_{args.gsp_prompts_per_group}_" f"gen_shared_prefix_{args.seed}_{args.gsp_num_groups}_{args.gsp_prompts_per_group}_"
f"{args.gsp_system_prompt_len}_{args.gsp_question_len}_{args.gsp_output_len}_" f"{args.gsp_system_prompt_len}_{args.gsp_question_len}_{args.gsp_output_len}_"
f"{tokenizer.__class__.__name__}.pkl" f"{tokenizer.__class__.__name__}.pkl"
) )