chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 16 Packages Projects Releases Wiki Activity

Revert "[AMD] support two batch overlapping for mori ep #17953" (#19161)

Browse Source
This commit is contained in:
Baizhou Zhang
2026-02-23 01:19:23 +08:00
committed by GitHub
parent 45095bac70
commit 43f83525c0
11 changed files with 157 additions and 725 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/test/bench_one_batch_server_internal.py
+5 -12
View File
@@ -787,24 +787,17 @@ def run_benchmark_internal(
else:
tokenizer = get_tokenizer(tokenizer_path)
# Get token capacity
internal_state = server_info.get("internal_states", [{}])
dp_size = internal_state[0].get("dp_size", None) or 1
skip_token_capacity_threshold = (
internal_state[0].get("memory_usage", {}).get("token_capacity", 1000000000)
)
# Get effective max running requests
max_running_requests_per_dp = internal_state[0].get(
"effective_max_running_requests_per_dp", -1
)
# Get token capacity
skip_token_capacity_threshold = 0
for i in range(dp_size):
skip_token_capacity_threshold += (
internal_state[i]
.get("memory_usage", {})
.get("token_capacity", 1000000000)
)
dp_size = server_info.get("dp_size", None) or 1
assert (
max_running_requests_per_dp > 0
), f"effective_max_running_requests_per_dp is not set, {max_running_requests_per_dp=}"