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

Revert "[CI] fix DS3.2 single node cudagraph_sizes config (#6241)" (#6497)

Browse Source 
# What this PR does / why we need it?
This PR reverts commit 8134146ab6, which
modified the DeepSeek V3.2 (W8A8) single-node nightly test
configuration. as there is no limit between tp_size and MTP.
# Does this PR introduce any user-facing change?
No. This PR only affects CI/CD test configurations and does not
introduce any user-facing changes.
# How was this patch tested?
N/A for a revert PR. The changes restore the previously known working
configuration.
- vLLM version: v0.15.0
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.15.0

Signed-off-by: guozr <guozr1997@hotmail.com>
Co-authored-by: guozr <guozr1997@hotmail.com>
This commit is contained in:
starmountain1997
2026-02-03 08:42:58 +08:00
committed by GitHub
parent c1618a0427
commit b6256e8bc9
1 changed files with 7 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
tests/e2e/nightly/single_node/models/test_deepseek_v3_2_w8a8.py
View File

@@ -54,7 +54,7 @@ aisbench_cases = [{
"max_out_len": 1500, "max_out_len": 1500,
"batch_size": 4, "batch_size": 4,
"request_rate": 11.2, "request_rate": 11.2,
"baseline": 110.5681, "baseline": 134,
"threshold": 0.97 "threshold": 0.97
}] }]
@@ -80,12 +80,13 @@ async def test_models(model: str, tp_size: int, dp_size: int) -> None:
str(tp_size), "--data-parallel-size", str(tp_size), "--data-parallel-size",
str(dp_size), "--port", str(dp_size), "--port",
str(port), "--max-model-len", "8192", "--max-num-batched-tokens", str(port), "--max-model-len", "8192", "--max-num-batched-tokens",
"8192", "--max-num-seqs", "8", "--trust-remote-code", "--quantization", "8192", "--max-num-seqs", "4", "--trust-remote-code", "--quantization",
"ascend", "--gpu-memory-utilization", "0.98", "--compilation-config", "ascend", "--gpu-memory-utilization", "0.92", "--compilation-config",
'{"cudagraph_capture_sizes":[8, 16, 24, 32], "cudagraph_mode":"FULL_DECODE_ONLY"}', '{"cudagraph_capture_sizes":[3, 6, 9, 12], "cudagraph_mode":"FULL_DECODE_ONLY"}',
"--speculative-config", "--speculative-config",
'{"num_speculative_tokens": 3, "method":"deepseek_mtp"}', '{"num_speculative_tokens": 2, "method":"deepseek_mtp"}',
"--additional-config", '{"layer_sharding": ["q_b_proj", "o_proj"]}', "--additional-config",
'{"layer_sharding": ["q_b_proj", "o_proj"]}',
"--reasoning-parser", "deepseek_v3", "--tokenizer_mode", "deepseek_v32" "--reasoning-parser", "deepseek_v3", "--tokenizer_mode", "deepseek_v32"
] ]
request_keyword_args: dict[str, Any] = { request_keyword_args: dict[str, Any] = {