EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix all cards super_pod_id same on A3 & proxy support min_tokens (#2939)

Browse Source 
### What this PR does / why we need it?
fix all cards super_pod_id same on A3 & proxy support min_tokens
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
2*A3 gen ranktable
before:
"prefill_device_list": [
        {
            "server_id": "xxx",
            "device_id": "0",
            "device_ip": "xxx",
            "super_pod_id": "0",
            "super_device_id": "106758159",
            "cluster_id": "1"
        },
        {
            "server_id": "xxx",
            "device_id": "1",
            "device_ip": "xxx",
            "super_pod_id": "0",
            "super_device_id": "106758159",
            "cluster_id": "2"
        }...
after:
"prefill_device_list": [
        {
            "server_id": "xxx",
            "device_id": "0",
            "device_ip": "xxx",
            "super_pod_id": "0",
            "super_device_id": "104857600",
            "cluster_id": "1"
        },
        {
            "server_id": "xxx",
            "device_id": "1",
            "device_ip": "xxx",
            "super_pod_id": "0",
            "super_device_id": "104923137",
            "cluster_id": "2"
        }...

---------

Signed-off-by: liziyu <liziyu16@huawei.com>
This commit is contained in:
liziyu
2025-09-16 01:09:18 +08:00
committed by GitHub
parent 382c29f3e1
commit aa3c4563ce
2 changed files with 3 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/disaggregated_prefill_v1/gen_ranktable.py
View File

@@ -73,6 +73,8 @@ if local_rank == "0":
super_pod_id = "0"
for idx in range(len(local_device_ids)):
device_id = local_device_ids[idx]
chip_id = device_id % chips_per_card
card_id = device_id // chips_per_card
if soc_info == AscendSocVersion.A3:
device_ip = get_cmd_stdout(
f"{hccn_tool_path} -i {device_id} -vnic -g | grep ipaddr"

1
examples/disaggregated_prefill_v1/load_balance_proxy_server_example.py
View File

@@ -363,6 +363,7 @@ async def send_request_to_service(client: httpx.AsyncClient,
}
req_data["stream"] = False
req_data["max_tokens"] = 1
req_data["min_tokens"] = 1
if "stream_options" in req_data:
del req_data["stream_options"]
headers = {